Add Image v2 as a sibling of Image edit, with separate Klein graphs.

Keep the v1 Image tab on its dual-branch template. The host agent follows Desktop's live port so Coolify stays on :8198.

Co-authored-by: Cursor <cursoragent@cursor.com>

.env.example

@@ -1,7 +1,7 @@
 PORT=3000
 HOST=0.0.0.0
-# ComfyUI on the Windows desktop. 8188 is often taken (SSH / other Comfy).
+# ComfyUI on the Windows desktop. Desktop hops ports (8188/8189/8190/...).
-# The host agent discovers the live instance (usually 8189) and proxies it on 8198.
+# The host agent (scripts/comfy-host-agent.mjs) finds the live instance and proxies it on 8198.
 COMFY_HOST=http://192.168.77.7:8198
 COMFY_PORT=8198
 # Optional: command used when this server runs on the Comfy host

docs/image-v2.md

@@ -0,0 +1,78 @@
+# Image v1 vs Image v2
+
+Image edit (v1) and Image v2 are siblings. v1 is unchanged. Do not mute-fix `workflow_flux2_klein_edit.json` (the dual-branch template that deletes `75:*` or `92:*` at runtime).
+
+| | Image edit (v1) | Image v2 |
+|---|---|---|
+| Tab | Image edit | Image v2 |
+| Route | `POST /api/edit` | `POST /api/v2/generate` and `POST /v2/generate` |
+| Graphs | One file, prune unused branch | `klein_v2_edit.json` or `klein_v2_compose.json` |
+| Second still | Optional; switches branch | Compose only; required. Mismatch is a 400 |
+| LoRAs | Optional user picker | SNOFS + Consistency, four strengths |
+| Defaults | 20 steps, CFG 1 | 24 steps, CFG 4 (turbo: 8 / 1) |
+
+Poll v2 jobs the same way as v1: `GET /api/generate/{id}/stream` and `GET /api/generate/{id}`.
+
+## Modes
+
+- **edit** — one image + text. Task is `scene`. Sending `image_b` is rejected.
+- **compose** — two images + text. `image_b` is required. No silent one-image fallback.
+
+Compose tasks:
+
+- **identity** — person from A; B reinforces the face; prompt changes pose/scene
+- **outfit** — person, body, pose, background from A; clothing only from B
+- **face_lock** — body/pose/scene from A; face from B
+- **scene** — A is the edit image; B is style/background
+
+Dummy test: Compose + two unrelated photos + “keep everything the same” must change the output. If it matches old v1 gens, still B is not connected.
+
+## Nodes the mapper patches
+
+Both graphs:
+
+| Node | Role |
+|------|------|
+| `1` | Load Image A (`inputs.image`) |
+| `2` / `23` | Scale to MP, lanczos (`megapixels`) |
+| `7` | SNOFS LoRA (`lora_name`, `strength_model`, `strength_clip`) |
+| `8` | Consistency LoRA (`lora_name`, `strength_model`, `strength_clip`) |
+| `9` | Positive `CLIPTextEncode.text` (role header + user prompt) |
+| `10` | Negative `CLIPTextEncode.text` |
+| `15` | Seed |
+| `17` | Steps (Flux2Scheduler; Klein-native, euler sampler) |
+| `18` | CFG |
+| `21` | SaveImage prefix |
+
+Compose only:
+
+| Node | Role |
+|------|------|
+| `22` | Load Image B |
+| `23` | Scale B |
+| `24` | VAE encode B |
+| `25` / `26` | Second ReferenceLatent (B fused into pos/neg) |
+
+If `image_b` is sent and the executed graph has fewer than two `LoadImage` nodes, the job fails.
+
+## Default sliders
+
+| Slider | Default |
+|--------|---------|
+| SNOFS Model | 0.65 |
+| SNOFS CLIP | 0.35 |
+| Consistency Model | 0.70 |
+| Consistency CLIP | 0.70 |
+| Steps | 24 (turbo 8) |
+| CFG | 4 (turbo 1) |
+| Scale to MP | 1.0 lanczos |
+
+There is no Denoise slider. This is reference-latent Klein, not inpaint.
+
+### NSFW identity vs outfit swap
+
+- **identity / face_lock** — keep Consistency at 0.70 / 0.70 so the face from A (identity) or B (face_lock) holds. Do not drop SNOFS CLIP below ~0.30 or the skin/body read falls apart.
+- **outfit** — keep SNOFS Model ~0.65 so cloth reads; do not raise it past ~0.85 or it starts rewriting the body from A. Consistency stays at 0.70 so the person in A does not become the person in B.
+- **scene / edit** — start at the table defaults. Turbo is for drafts only.
+
+Presets on this tab store sliders + mode + task only. They do not load v1 LoRA stacks or a cached latent.

pages/index.vue

@@ -54,9 +54,21 @@
             >
               Image edit
             </button>
+            <button
+              type="button"
+              class="rounded-full border px-3 py-1 text-xs font-medium"
+              :class="studioMode === 'editv2' ? 'border-amber-300/70 bg-amber-400/10 text-amber-100' : 'border-white/10 text-zinc-400 hover:text-white'"
+              @click="studioMode = 'editv2'; outputFocus = 'edit'"
+            >
+              Image v2
+            </button>
           </div>
           <h2 class="font-display text-2xl font-bold">Input</h2>
-          <p class="text-sm text-zinc-400">{{ studioMode === 'edit'
+          <p class="text-sm text-zinc-400">{{ studioMode === 'editv2'
+            ? (v2Mode === 'compose'
+              ? 'Compose: still A is the person/body. Still B is the face or outfit. Two stills required — no silent one-image fallback.'
+              : 'Edit: one still plus text. Compose is a separate graph if you need a second still.')
+            : studioMode === 'edit'
             ? 'Drop the still to edit. An optional second still applies image 2 onto image 1. Files leave the desktop after save.'
             : 'Drop a still, describe the motion, then send the job. Generate stays available — a long shot list takes one queue slot, and you can line up more while it runs.' }}</p>
         </div>
@@ -80,14 +92,45 @@
             <button type="button" class="relative mt-3 rounded-full bg-zinc-950/80 px-3 py-1 text-xs" @click.stop="resetImage">Reset</button>
           </div>
           <div v-else class="flex min-h-44 flex-col items-center justify-center text-center">
-            <p class="font-medium">{{ studioMode === 'edit' ? 'Image 1' : (textToVideo ? 'Start still · optional' : 'Choose an image') }}</p>
+            <p class="font-medium">{{ studioMode === 'edit' || studioMode === 'editv2' ? (studioMode === 'editv2' ? 'Still A' : 'Image 1') : (textToVideo ? 'Start still · optional' : 'Choose an image') }}</p>
             <p class="mt-1 text-sm text-zinc-500">{{ textToVideo ? 'Text-to-video does not need a still. Shot 2+ will still extend from the last frame.' : 'Browse the library or upload a new PNG, JPG, or WEBP' }}</p>
           </div>
         </div>
 
-        <div v-if="studioMode === 'edit'" class="rounded-2xl border border-dashed border-white/15 bg-zinc-950/50 p-4">
+        <div v-if="studioMode === 'editv2'" class="flex flex-wrap gap-2">
+          <button
+            type="button"
+            class="rounded-full border px-3 py-1 text-xs font-medium"
+            :class="v2Mode === 'edit' ? 'border-amber-300/70 bg-amber-400/10 text-amber-100' : 'border-white/10 text-zinc-400 hover:text-white'"
+            @click="v2Mode = 'edit'; v2Task = 'scene'"
+          >
+            Edit
+          </button>
+          <button
+            type="button"
+            class="rounded-full border px-3 py-1 text-xs font-medium"
+            :class="v2Mode === 'compose' ? 'border-amber-300/70 bg-amber-400/10 text-amber-100' : 'border-white/10 text-zinc-400 hover:text-white'"
+            @click="v2Mode = 'compose'"
+          >
+            Compose
+          </button>
+          <label v-if="v2Mode === 'compose'" class="ml-auto text-xs">
+            <span class="mr-2 text-zinc-500">Task</span>
+            <select
+              v-model="v2Task"
+              class="rounded-lg border border-white/10 bg-zinc-950 px-2 py-1 text-xs text-zinc-300 outline-none ring-amber-300/40 focus:ring-2"
+            >
+              <option value="scene">scene</option>
+              <option value="identity">identity</option>
+              <option value="outfit">outfit</option>
+              <option value="face_lock">face_lock</option>
+            </select>
+          </label>
+        </div>
+
+        <div v-if="studioMode === 'edit' || (studioMode === 'editv2' && v2Mode === 'compose')" class="rounded-2xl border border-dashed border-white/15 bg-zinc-950/50 p-4">
           <div class="mb-2 flex items-center justify-between gap-2">
-            <p class="text-sm font-medium text-zinc-300">Image 2 · optional</p>
+            <p class="text-sm font-medium text-zinc-300">{{ studioMode === 'editv2' ? 'Still B · required' : 'Image 2 · optional' }}</p>
             <button
               v-if="editRefFile"
               type="button"
@@ -116,19 +159,21 @@
               class="max-h-40 w-full object-contain bg-black/40"
             >
             <span v-else class="flex min-h-28 items-center justify-center px-3 py-6 text-center text-sm text-zinc-500">
-              Apply this still onto image 1
+              {{ studioMode === 'editv2' ? stillBHint : 'Apply this still onto image 1' }}
             </span>
           </button>
           <p class="mt-2 text-xs text-zinc-500">
-            {{ editRefFile
+            {{ studioMode === 'editv2'
+              ? (editRefFile ? stillBHint : 'Compose will reject this job until still B is loaded.')
+              : (editRefFile
                 ? 'Two-image branch on. Image 2 is applied onto image 1.'
-              : 'One-image branch on. Add a second still to transfer details from image 2.' }}
+                : 'One-image branch on. Add a second still to transfer details from image 2.') }}
           </p>
         </div>
 
         <div>
           <div class="mb-2 flex flex-wrap items-center justify-between gap-2">
-            <label class="text-sm font-medium text-zinc-300">{{ studioMode === 'edit' ? 'Edit prompt' : 'Motion & scene prompt' }}</label>
+            <label class="text-sm font-medium text-zinc-300">{{ studioMode === 'edit' || studioMode === 'editv2' ? 'Edit prompt' : 'Motion & scene prompt' }}</label>
             <div class="flex flex-wrap items-center gap-3">
               <button
                 v-if="identityPrompting"
@@ -175,7 +220,7 @@
           </p>
           <div
             class="grid gap-3"
-            :class="studioMode === 'edit' ? 'md:grid-cols-2' : ''"
+            :class="studioMode === 'edit' || studioMode === 'editv2' ? 'md:grid-cols-2' : ''"
           >
             <div class="min-w-0">
               <textarea
@@ -194,9 +239,17 @@
                 <span class="mt-2 inline-block text-xs text-amber-200">{{ prompt.trim() ? 'View / edit prompt' : 'Write prompt' }}</span>
               </button>
             </div>
-            <label v-if="studioMode === 'edit'" class="block min-w-0 text-sm">
+            <label v-if="studioMode === 'edit' || studioMode === 'editv2'" class="block min-w-0 text-sm">
               <span class="mb-2 block font-medium text-zinc-300">Negative prompt</span>
               <textarea
+                v-if="studioMode === 'editv2'"
+                v-model="v2Negative"
+                rows="2"
+                class="min-h-[4.5rem] w-full rounded-2xl border border-white/10 bg-zinc-950/70 px-4 py-3 text-sm outline-none ring-amber-300/40 placeholder:text-zinc-600 [overflow-wrap:anywhere] focus:ring-2 md:h-[calc(100%-1.75rem)] md:min-h-[8.5rem]"
+                placeholder="What to avoid…"
+              />
+              <textarea
+                v-else
                 v-model="editNegative"
                 rows="2"
                 class="min-h-[4.5rem] w-full rounded-2xl border border-white/10 bg-zinc-950/70 px-4 py-3 text-sm outline-none ring-amber-300/40 placeholder:text-zinc-600 [overflow-wrap:anywhere] focus:ring-2 md:h-[calc(100%-1.75rem)] md:min-h-[8.5rem]"
@@ -617,7 +670,62 @@
             @remove="deleteGenerationPreset"
           />
         </div>
-        <div v-else class="space-y-3">
+        <div v-else-if="studioMode === 'editv2'" class="space-y-3">
+          <p class="text-xs text-zinc-500">Klein v2 · 9B Base · euler. Edit and Compose are separate graphs. Shares the desktop GPU with video.</p>
+          <div class="grid grid-cols-2 gap-3">
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">SNOFS Model</span>
+              <input v-model.number="v2SnofsModel" type="number" min="0" max="2" step="0.05" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">SNOFS CLIP</span>
+              <input v-model.number="v2SnofsClip" type="number" min="0" max="2" step="0.05" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">Consistency Model</span>
+              <input v-model.number="v2ConsistencyModel" type="number" min="0" max="2" step="0.05" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">Consistency CLIP</span>
+              <input v-model.number="v2ConsistencyClip" type="number" min="0" max="2" step="0.05" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">Steps</span>
+              <input v-model.number="v2Steps" type="number" :min="IMAGE_STEPS_MIN" :max="IMAGE_STEPS_MAX" step="1" :disabled="v2Turbo" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2 disabled:opacity-50">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">CFG</span>
+              <input v-model.number="v2Cfg" type="number" :min="IMAGE_CFG_MIN" :max="IMAGE_CFG_MAX" :step="IMAGE_CFG_STEP" :disabled="v2Turbo" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2 disabled:opacity-50">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">Seed</span>
+              <input v-model="seedInput" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm" placeholder="random">
+            </label>
+            <label class="block text-sm">
+              <span class="mb-1 block text-zinc-400">Scale to MP</span>
+              <input v-model.number="v2Megapixels" type="number" :min="IMAGE_SCALE_MP_MIN" :max="IMAGE_SCALE_MP_MAX" :step="IMAGE_SCALE_MP_STEP" class="w-full rounded-xl border border-white/10 bg-zinc-950 px-3 py-2 text-sm outline-none ring-amber-300/40 focus:ring-2">
+            </label>
+          </div>
+          <label class="flex cursor-pointer items-center gap-2 text-xs text-zinc-400">
+            <span class="relative inline-flex h-5 w-9 shrink-0 items-center">
+              <input v-model="v2Turbo" type="checkbox" class="peer sr-only" role="switch">
+              <span class="absolute inset-0 rounded-full bg-zinc-700 transition peer-checked:bg-amber-400" />
+              <span class="absolute left-0.5 top-0.5 h-4 w-4 rounded-full bg-white transition peer-checked:translate-x-4" />
+            </span>
+            Turbo · 8 steps, CFG 1
+          </label>
+          <div class="max-w-md">
+            <GenerationPresetBar
+              :presets="imageV2GenerationPresets"
+              v-model:selected-id="imageV2PresetId"
+              v-model:name="imageV2PresetName"
+              @load="loadGenerationPreset"
+              @save="saveGenerationPreset"
+              @remove="deleteGenerationPreset"
+            />
+          </div>
+        </div>
+        <div v-else-if="studioMode === 'edit'" class="space-y-3">
           <p class="text-xs text-zinc-500">Flux.2 Klein 9B. Shares the desktop GPU with video.</p>
           <div class="max-w-md space-y-2">
             <LoraSelect v-model="imageLoraStack" :names="imageLoraOptions" label="LoRA" empty-label="None" compact />
@@ -795,10 +903,10 @@
           <button
             type="button"
             class="flex-1 rounded-2xl bg-amber-400 px-5 py-3 font-semibold text-zinc-950 transition hover:bg-amber-300 disabled:cursor-not-allowed disabled:opacity-50"
-            :disabled="studioMode === 'edit' ? editBlocked : videoGenerateDisabled"
+            :disabled="studioMode === 'editv2' ? editV2Blocked : (studioMode === 'edit' ? editBlocked : videoGenerateDisabled)"
-            @click="studioMode === 'edit' ? editImage() : generate()"
+            @click="studioMode === 'editv2' ? editImageV2() : (studioMode === 'edit' ? editImage() : generate())"
           >
-            {{ studioMode === 'edit' ? editSubmitLabel : generateLabel }}
+            {{ studioMode === 'editv2' ? editV2SubmitLabel : (studioMode === 'edit' ? editSubmitLabel : generateLabel) }}
           </button>
           <button
             v-if="videoBusy || editBusy"
@@ -810,9 +918,10 @@
           </button>
         </div>
         <p v-if="studioMode === 'video' && videoBusy" class="text-xs text-zinc-400">The current job keeps running in Output. This submit waits in the job queue.</p>
-        <p v-else-if="studioMode === 'edit' && (videoBusy || editBusy)" class="text-xs text-zinc-400">The current job keeps running in Output. This submit waits in the job queue.</p>
+        <p v-else-if="(studioMode === 'edit' || studioMode === 'editv2') && (videoBusy || editBusy)" class="text-xs text-zinc-400">The current job keeps running in Output. This submit waits in the job queue.</p>
         <p v-if="studioMode === 'video' && videoBlockReason" class="text-xs text-zinc-400">{{ videoBlockReason }}</p>
         <p v-if="studioMode === 'edit' && editBlockReason" class="text-xs text-zinc-400">{{ editBlockReason }}</p>
+        <p v-if="studioMode === 'editv2' && editV2BlockReason" class="text-xs text-zinc-400">{{ editV2BlockReason }}</p>
 
         <div v-if="studioJobs.length" class="rounded-2xl border border-white/10 bg-zinc-950/40 p-4">
           <div class="flex items-start justify-between gap-3">
@@ -836,7 +945,7 @@
               <div class="min-w-0 flex-1">
                 <p class="truncate text-sm font-medium text-zinc-100">{{ job.name }}</p>
                 <p class="mt-0.5 text-xs text-zinc-500">
-                  <span v-if="job.kind === 'edit'" class="mr-1 text-amber-200/80">{{ (job.shotCount || 1) > 1 ? 'EDIT' : 'IMAGE' }} ·</span>
+                  <span v-if="job.kind === 'edit'" class="mr-1 text-amber-200/80">{{ job.imagePipeline === 'v2' ? 'IMAGE V2' : ((job.shotCount || 1) > 1 ? 'EDIT' : 'IMAGE') }} ·</span>
                   {{ studioJobStatusLabel(job) }}
                   · {{ studioJobShotLine(job) }}
                 </p>
@@ -1545,7 +1654,7 @@
     <PromptDrawer
       :open="promptDrawerOpen"
       v-model:text="prompt"
-      :title="studioMode === 'edit' ? 'Edit prompt' : (shotScriptMode ? 'Shot script' : 'Prompt')"
+      :title="studioMode === 'edit' || studioMode === 'editv2' ? 'Edit prompt' : (shotScriptMode ? 'Shot script' : 'Prompt')"
       :placeholder="promptPlaceholder"
       @update:open="promptDrawerOpen = $event"
     />
@@ -1671,6 +1780,18 @@ import {
   VIDEO_CFG_STEP,
   type GenerationPreset
 } from '~/utils/generationPresets'
+import {
+  IMAGE_V2_CFG_DEFAULT,
+  IMAGE_V2_CONSISTENCY_CLIP,
+  IMAGE_V2_CONSISTENCY_MODEL,
+  IMAGE_V2_SNOFS_CLIP,
+  IMAGE_V2_SNOFS_MODEL,
+  IMAGE_V2_STEPS_DEFAULT,
+  clampImageV2Strength,
+  type ImageV2Mode,
+  type ImageV2PresetSettings,
+  type ImageV2Task
+} from '~/utils/imageV2'
 import {
   coerceVideoWorkflow,
   composeVideoWorkflow,
@@ -1863,10 +1984,13 @@ const shotLoraStacks = ref<Record<number, LoraStackItem[]>>({})
 const extendLoraStack = ref<LoraStackItem[]>([])
 const videoGenerationPresets = ref<GenerationPreset[]>([])
 const imageGenerationPresets = ref<GenerationPreset[]>([])
+const imageV2GenerationPresets = ref<GenerationPreset[]>([])
 const videoPresetId = ref('')
 const imagePresetId = ref('')
+const imageV2PresetId = ref('')
 const videoPresetName = ref('')
 const imagePresetName = ref('')
+const imageV2PresetName = ref('')
 const CFG_MIN = VIDEO_CFG_MIN
 const CFG_MAX = VIDEO_CFG_MAX
 const CFG_STEP = VIDEO_CFG_STEP
@@ -1933,6 +2057,17 @@ const editScaleToTotalPixels = ref(false)
 const editScaleMegapixels = ref(1)
 const editRefFile = ref<File | null>(null)
 const editRefPreview = ref('')
+const v2Mode = ref<ImageV2Mode>('edit')
+const v2Task = ref<ImageV2Task>('scene')
+const v2Negative = ref('')
+const v2SnofsModel = ref(IMAGE_V2_SNOFS_MODEL)
+const v2SnofsClip = ref(IMAGE_V2_SNOFS_CLIP)
+const v2ConsistencyModel = ref(IMAGE_V2_CONSISTENCY_MODEL)
+const v2ConsistencyClip = ref(IMAGE_V2_CONSISTENCY_CLIP)
+const v2Steps = ref(IMAGE_V2_STEPS_DEFAULT)
+const v2Cfg = ref(IMAGE_V2_CFG_DEFAULT)
+const v2Megapixels = ref(1)
+const v2Turbo = ref(false)
 const editChainStep = ref(1)
 const editChainTotal = ref(1)
 const editChainLabel = ref('')
@@ -2010,6 +2145,7 @@ const studioJobs = ref<Array<{
   lastError?: string
   completedCount?: number
   pendingCount?: number
+  imagePipeline?: string
 }>>([])
 const outputStudioJobId = ref('')
 const videoAwaitingBeast = ref(false)
@@ -2044,8 +2180,8 @@ const LOCKS_STORE = 'aigen-global-locks'
 const REFS_STORE = 'aigen-permanence-refs'
 const pickerHeading = computed(() => {
   if (pickerKind.value === 'permanence') return 'Choose a still to name'
-  if (pickerSlot.value === 'main') return studioMode.value === 'edit' ? 'Choose image 1' : 'Choose an image'
+  if (pickerSlot.value === 'main') return studioMode.value === 'edit' || studioMode.value === 'editv2' ? (studioMode.value === 'editv2' ? 'Choose still A' : 'Choose image 1') : 'Choose an image'
-  if (pickerSlot.value === 'editRef') return 'Choose image 2'
+  if (pickerSlot.value === 'editRef') return studioMode.value === 'editv2' ? 'Choose still B' : 'Choose image 2'
   return `Choose Picture ${Number(pickerSlot.value) + 2}`
 })
 const permanencePicker = computed(() => pickerKind.value === 'permanence')
@@ -2299,6 +2435,32 @@ const editBlockReason = computed(() => {
   if (!comfyOk.value && !imageComfyOk.value) return 'Desktop GPU is offline.'
   return ''
 })
+const stillBHint = computed(() => {
+  if (v2Task.value === 'outfit') return 'Clothing only from still B. Person, body, pose, and background stay on still A.'
+  if (v2Task.value === 'face_lock') return 'Exact face from still B. Body, pose, and scene stay on still A.'
+  if (v2Task.value === 'identity') return 'Still B reinforces the face. Still A is the person.'
+  return 'Still A is the edit image. Still B is style or background.'
+})
+const editV2Blocked = computed(() => {
+  if (!file.value || !prompt.value.trim() || !folderId.value) return true
+  if (v2Mode.value === 'compose' && !editRefFile.value) return true
+  if (!comfyOk.value && !imageComfyOk.value) return true
+  return false
+})
+const editV2BlockReason = computed(() => {
+  if (!file.value) return 'Load still A first.'
+  if (v2Mode.value === 'compose' && !editRefFile.value) return 'Compose requires still B. This will not fall back to one-image generation.'
+  if (!prompt.value.trim()) return 'Write a prompt first.'
+  if (!folderId.value) return 'Choose a library folder before generating.'
+  if (!imageComfyConfigured.value && !comfyOk.value) return 'Image v2 is not configured. Set COMFY_HOST.'
+  if (!comfyOk.value && !imageComfyOk.value) return 'Desktop GPU is offline.'
+  return ''
+})
+const editV2SubmitLabel = computed(() => {
+  const occupied = videoBusy.value || editBusy.value || studioJobs.value.some(job => job.status === 'running')
+  const label = v2Mode.value === 'compose' ? 'Compose image' : 'Edit image'
+  return occupied ? `Queue ${label.toLowerCase()}` : label
+})
 const generateLabel = computed(() => {
   const n = queuedExtensionCount.value
   if (videoBusy.value) return n ? `Queue ${n + 1} shots` : 'Queue job'
@@ -2353,6 +2515,12 @@ const composedIdentityPrompt = computed(() => {
   return buildIdentityPrompt(identityActionPrompt.value, extraIdentityPictures.value)
 })
 const promptPlaceholder = computed(() => {
+  if (studioMode.value === 'editv2') {
+    if (v2Mode.value === 'compose' && v2Task.value === 'outfit') return 'Keep everything the same except the clothes from still B.'
+    if (v2Mode.value === 'compose' && v2Task.value === 'face_lock') return 'Keep the body and scene from still A. Use the face from still B.'
+    if (v2Mode.value === 'compose' && v2Task.value === 'identity') return 'Same person as still A. Change the pose and scene.'
+    return 'Make the lighting warmer. Keep the face unchanged.'
+  }
   if (studioMode.value === 'edit') {
     return editRefFile.value
       ? 'Apply the dress in image 2 to the person in image 1. Preserve the exact features, proportions, pose, and setting of image 1.'
@@ -3064,18 +3232,41 @@ function applyLoraSelection(stack?: unknown, shotStacks?: Array<unknown>) {
   shotLoraStacks.value = next
 }
 
-const currentPresetKind = computed(() => studioMode.value === 'edit' ? 'image' : 'video')
+const currentPresetKind = computed(() => (
+  studioMode.value === 'editv2' ? 'imagev2' : studioMode.value === 'edit' ? 'image' : 'video'
+))
 
 async function loadGenerationPresets() {
-  const [video, image] = await Promise.all([
+  const [video, image, imagev2] = await Promise.all([
     $fetch<{ presets?: GenerationPreset[] }>('/api/generation-presets', { query: { kind: 'video' } }).catch(() => ({ presets: [] })),
-    $fetch<{ presets?: GenerationPreset[] }>('/api/generation-presets', { query: { kind: 'image' } }).catch(() => ({ presets: [] }))
+    $fetch<{ presets?: GenerationPreset[] }>('/api/generation-presets', { query: { kind: 'image' } }).catch(() => ({ presets: [] })),
+    $fetch<{ presets?: GenerationPreset[] }>('/api/generation-presets', { query: { kind: 'imagev2' } }).catch(() => ({ presets: [] }))
   ])
   videoGenerationPresets.value = video.presets || []
   imageGenerationPresets.value = image.presets || []
+  imageV2GenerationPresets.value = imagev2.presets || []
 }
 
 function currentPresetSnapshot() {
+  if (studioMode.value === 'editv2') {
+    return {
+      kind: 'imagev2' as const,
+      loraStack: [],
+      settings: {
+        mode: v2Mode.value,
+        task: v2Mode.value === 'compose' ? v2Task.value : 'scene',
+        negative: v2Negative.value,
+        snofsModel: clampImageV2Strength(v2SnofsModel.value, IMAGE_V2_SNOFS_MODEL),
+        snofsClip: clampImageV2Strength(v2SnofsClip.value, IMAGE_V2_SNOFS_CLIP),
+        consistencyModel: clampImageV2Strength(v2ConsistencyModel.value, IMAGE_V2_CONSISTENCY_MODEL),
+        consistencyClip: clampImageV2Strength(v2ConsistencyClip.value, IMAGE_V2_CONSISTENCY_CLIP),
+        steps: clampImageSteps(v2Steps.value, IMAGE_V2_STEPS_DEFAULT),
+        cfg: clampImageCfg(v2Cfg.value, IMAGE_V2_CFG_DEFAULT),
+        megapixels: clampImageScaleMegapixels(v2Megapixels.value, 1),
+        turbo: v2Turbo.value === true
+      }
+    }
+  }
   if (studioMode.value === 'edit') {
     return {
       kind: 'image' as const,
@@ -3109,7 +3300,22 @@ function currentPresetSnapshot() {
 function applyGenerationPreset(preset: GenerationPreset) {
   const stack = filterLoraStackForStudio(preset.loraStack, ltxEnabled.value)
   const skipped = preset.loraStack.length - stack.length
-  if (preset.kind === 'image') {
+  if (preset.kind === 'imagev2') {
+    const settings = preset.settings as ImageV2PresetSettings
+    v2Mode.value = settings.mode === 'compose' ? 'compose' : 'edit'
+    v2Task.value = v2Mode.value === 'compose' ? (settings.task || 'scene') : 'scene'
+    if (typeof settings.negative === 'string') v2Negative.value = settings.negative
+    v2SnofsModel.value = clampImageV2Strength(settings.snofsModel, IMAGE_V2_SNOFS_MODEL)
+    v2SnofsClip.value = clampImageV2Strength(settings.snofsClip, IMAGE_V2_SNOFS_CLIP)
+    v2ConsistencyModel.value = clampImageV2Strength(settings.consistencyModel, IMAGE_V2_CONSISTENCY_MODEL)
+    v2ConsistencyClip.value = clampImageV2Strength(settings.consistencyClip, IMAGE_V2_CONSISTENCY_CLIP)
+    v2Steps.value = clampImageSteps(settings.steps, IMAGE_V2_STEPS_DEFAULT)
+    v2Cfg.value = clampImageCfg(settings.cfg, IMAGE_V2_CFG_DEFAULT)
+    v2Megapixels.value = clampImageScaleMegapixels(settings.megapixels, 1)
+    v2Turbo.value = settings.turbo === true
+    imageV2PresetId.value = preset.id
+    imageV2PresetName.value = preset.name
+  } else if (preset.kind === 'image') {
     imageLoraStack.value = stack
     const settings = preset.settings as {
       steps?: number
@@ -3168,20 +3374,31 @@ function applyGenerationPreset(preset: GenerationPreset) {
   })
 }
 
+function currentPresetList() {
+  if (currentPresetKind.value === 'imagev2') return imageV2GenerationPresets.value
+  if (currentPresetKind.value === 'image') return imageGenerationPresets.value
+  return videoGenerationPresets.value
+}
+
 function loadGenerationPreset(id: string) {
-  const list = currentPresetKind.value === 'image' ? imageGenerationPresets.value : videoGenerationPresets.value
+  const preset = currentPresetList().find(item => item.id === id)
-  const preset = list.find(item => item.id === id)
   if (!preset) return
   applyGenerationPreset(preset)
 }
 
 async function saveGenerationPreset() {
-  const name = (currentPresetKind.value === 'image' ? imagePresetName.value : videoPresetName.value).trim()
+  const name = (
+    currentPresetKind.value === 'imagev2'
+      ? imageV2PresetName.value
+      : currentPresetKind.value === 'image'
+        ? imagePresetName.value
+        : videoPresetName.value
+  ).trim()
   if (!name) {
     toast('Name the preset before saving.')
     return
   }
-  const list = currentPresetKind.value === 'image' ? imageGenerationPresets.value : videoGenerationPresets.value
+  const list = currentPresetList()
   const existing = list.find(item => item.name.toLowerCase() === name.toLowerCase())
   if (existing && !window.confirm(`Overwrite preset “${existing.name}”?`)) return
   try {
@@ -3190,7 +3407,11 @@ async function saveGenerationPreset() {
       method: 'POST',
       body
     })
-    if (saved.preset.kind === 'image') {
+    if (saved.preset.kind === 'imagev2') {
+      imageV2GenerationPresets.value = saved.presets
+      imageV2PresetId.value = saved.preset.id
+      imageV2PresetName.value = saved.preset.name
+    } else if (saved.preset.kind === 'image') {
       imageGenerationPresets.value = saved.presets
       imagePresetId.value = saved.preset.id
       imagePresetName.value = saved.preset.name
@@ -3206,8 +3427,12 @@ async function saveGenerationPreset() {
 }
 
 async function deleteGenerationPreset() {
-  const id = currentPresetKind.value === 'image' ? imagePresetId.value : videoPresetId.value
+  const id = currentPresetKind.value === 'imagev2'
-  const list = currentPresetKind.value === 'image' ? imageGenerationPresets.value : videoGenerationPresets.value
+    ? imageV2PresetId.value
+    : currentPresetKind.value === 'image'
+      ? imagePresetId.value
+      : videoPresetId.value
+  const list = currentPresetList()
   const preset = list.find(item => item.id === id)
   if (!id || !preset) return
   if (!window.confirm(`Delete preset “${preset.name}”?`)) return
@@ -3216,7 +3441,10 @@ async function deleteGenerationPreset() {
       method: 'DELETE',
       query: { kind: currentPresetKind.value }
     })
-    if (currentPresetKind.value === 'image') {
+    if (currentPresetKind.value === 'imagev2') {
+      imageV2GenerationPresets.value = saved.presets
+      imageV2PresetId.value = ''
+    } else if (currentPresetKind.value === 'image') {
       imageGenerationPresets.value = saved.presets
       imagePresetId.value = ''
     } else {
@@ -4869,6 +5097,91 @@ async function editImage() {
   }
 }
 
+async function editImageV2() {
+  if (editV2Blocked.value) {
+    toast(editV2BlockReason.value || 'Load still A first.')
+    return
+  }
+  if (v2Mode.value === 'compose' && !editRefFile.value) {
+    toast('Compose requires still B. This will not fall back to one-image generation.')
+    return
+  }
+  const hideOut = hideThumbnail.value
+  try {
+    const body = new FormData()
+    body.append('mode', v2Mode.value)
+    body.append('task', v2Mode.value === 'compose' ? v2Task.value : 'scene')
+    body.append('image_a', file.value as File)
+    if (v2Mode.value === 'compose' && editRefFile.value) body.append('image_b', editRefFile.value)
+    body.append('prompt', prompt.value.trim())
+    body.append('negative', v2Negative.value)
+    body.append('snofs_model', String(clampImageV2Strength(v2SnofsModel.value, IMAGE_V2_SNOFS_MODEL)))
+    body.append('snofs_clip', String(clampImageV2Strength(v2SnofsClip.value, IMAGE_V2_SNOFS_CLIP)))
+    body.append('consistency_model', String(clampImageV2Strength(v2ConsistencyModel.value, IMAGE_V2_CONSISTENCY_MODEL)))
+    body.append('consistency_clip', String(clampImageV2Strength(v2ConsistencyClip.value, IMAGE_V2_CONSISTENCY_CLIP)))
+    body.append('steps', String(clampImageSteps(v2Steps.value, IMAGE_V2_STEPS_DEFAULT)))
+    body.append('cfg', String(clampImageCfg(v2Cfg.value, IMAGE_V2_CFG_DEFAULT)))
+    body.append('megapixels', String(clampImageScaleMegapixels(v2Megapixels.value, 1)))
+    body.append('turbo', String(v2Turbo.value === true))
+    body.append('seed', seedInput.value || 'random')
+    body.append('folderId', folderId.value)
+    body.append('hideInput', String(showPrivacyToggles.value && hideInputPreview.value))
+    body.append('hideThumbnail', String(showPrivacyToggles.value && hideThumbnail.value))
+    body.append('name', clipName.value.trim())
+    const downloadName = clipFileName(clipName.value.trim() || prompt.value.split(/[.!?\n]/)[0] || 'aigen-v2')
+    const started = await $fetch<{
+      jobId: string
+      studioJobId?: string
+      queued?: boolean
+      steps: number
+      hideThumbnail?: boolean
+      folderLocked?: boolean
+    }>('/api/v2/generate', {
+      method: 'POST',
+      body
+    })
+    if (started.queued) {
+      toast('Queued Klein v2 behind the current job.')
+      await refreshStudioQueue()
+      return
+    }
+    editStatusBusy.value = false
+    editBusy.value = true
+    stopListen('edit')
+    editSettledUi = false
+    editProgress.value = 2
+    editResultUrl.value = ''
+    editAwaitingReveal.value = false
+    concealEditOutput.value = hideOut
+    editLockedSave.value = false
+    editStatusMessage.value = 'Checking Beast ComfyUI...'
+    editMaxStep.value = started.steps
+    editChainStep.value = 1
+    editChainTotal.value = 1
+    editChainLabel.value = ''
+    editOverallProgress.value = 0
+    editCompletedChainStep.value = 0
+    editActiveChainPlan.value = [{ label: v2Mode.value === 'compose' ? 'Compose' : 'Edit', prompt: prompt.value.trim() }]
+    startTimer('edit')
+    editDownloadName.value = downloadName
+    editJobId.value = started.jobId
+    outputStudioJobId.value = started.studioJobId || ''
+    persistActiveJob('edit', started.jobId, started.hideThumbnail === true || hideOut, started.folderLocked === true)
+    listen('edit', started.jobId, started.hideThumbnail === true || hideOut, started.folderLocked === true)
+    void pollComfyHealth()
+    void refreshStudioQueue()
+  } catch (error: any) {
+    editSettledUi = true
+    const message = error?.data?.statusMessage || error?.statusMessage || error?.message || 'Image v2 failed'
+    editStatusMessage.value = message
+    toast(message)
+    editBusy.value = false
+    editStatusBusy.value = false
+    stopTimer('edit')
+    void pollComfyHealth()
+  }
+}
+
 async function useEditAsInput() {
   if (!editResultUrl.value) return
   try {

pages/queue.vue

@@ -457,6 +457,7 @@ type StudioJobRow = {
   shots?: Queue | null
   plannedShots?: { prompt: string; duration: number; loraName?: string; loraStack?: LoraStackItem[] }[]
   completedCount?: number
+  imagePipeline?: string
 }
 
 type PendingRemove = {
@@ -759,7 +760,10 @@ function listenJob(job: StudioJobRow) {
 }
 
 function jobKindLabel(job: StudioJobRow) {
-  if (job.kind === 'edit') return (job.shotCount || 1) > 1 ? 'EDIT' : 'IMAGE'
+  if (job.kind === 'edit') {
+    if (job.imagePipeline === 'v2') return 'IMAGE V2'
+    return (job.shotCount || 1) > 1 ? 'EDIT' : 'IMAGE'
+  }
   return 'Job'
 }
 

scripts/comfy-host-agent.mjs

@@ -23,36 +23,73 @@ function authorized(req) {
   return header === `Bearer ${token}`
 }
 
-function candidatePorts() {
+function lockPorts() {
-  const ports = new Set([8188, 8189])
+  const ports = []
-  try {
-    const url = new URL(defaultHttp.includes('://') ? defaultHttp : `http://${defaultHttp}`)
-    if (url.port) ports.add(Number(url.port))
-  } catch { /* ignore */ }
   try {
     const dir = join(process.env.APPDATA || '', 'Comfy Desktop', 'port-locks')
     for (const name of readdirSync(dir)) {
       const match = name.match(/^port-(\d+)\.json$/)
-      if (match) ports.add(Number(match[1]))
+      if (match) ports.push(Number(match[1]))
     }
   } catch { /* ignore */ }
-  return [...ports]
+  return ports
 }
 
-async function probe(portNum) {
+function candidatePorts() {
+  const skip = new Set([port, proxyPort])
+  const ordered = []
+  const seen = new Set()
+  const add = (value) => {
+    const next = Number(value)
+    if (!Number.isInteger(next) || next < 1 || next > 65535) return
+    if (skip.has(next) || seen.has(next)) return
+    seen.add(next)
+    ordered.push(next)
+  }
+  for (const next of lockPorts()) add(next)
+  try {
+    const url = new URL(defaultHttp.includes('://') ? defaultHttp : `http://${defaultHttp}`)
+    if (url.port) add(url.port)
+  } catch { /* ignore */ }
+  for (let next = 8188; next <= 8210; next++) add(next)
+  return ordered
+}
+
+async function probeStats(portNum) {
   try {
     const res = await fetch(`http://127.0.0.1:${portNum}/system_stats`, { signal: AbortSignal.timeout(2500) })
-    return res.ok
+    if (!res.ok) return null
+    const stats = await res.json()
+    if (!stats?.system) return null
+    return stats
   } catch {
-    return false
+    return null
   }
 }
 
+function advertisedPort(stats) {
+  const argv = stats?.system?.argv
+  if (!Array.isArray(argv)) return 0
+  const index = argv.findIndex(item => String(item) === '--port')
+  if (index >= 0) return Number(argv[index + 1]) || 0
+  const flag = argv.find(item => /^--port=\d+$/.test(String(item)))
+  if (flag) return Number(String(flag).split('=')[1]) || 0
+  return 0
+}
+
 async function findHealthyPort() {
+  const found = []
   for (const next of candidatePorts()) {
-    if (await probe(next)) return next
+    const stats = await probeStats(next)
+    if (!stats) continue
+    found.push({ port: next, advertised: advertisedPort(stats) })
   }
-  return 0
+  const native = found.find(item => item.advertised && item.advertised === item.port)
+  if (native) return native.port
+  for (const item of found) {
+    if (item.advertised && found.some(other => other.port === item.advertised)) return item.advertised
+  }
+  return found[0]?.port || 0
 }
 
 let proxyServer = null
@@ -60,12 +97,11 @@ let proxyTarget = 0
 
 function ensureProxy(targetPort) {
   if (!targetPort) return
-  if (proxyServer && proxyTarget === targetPort) return
+  if (proxyTarget !== targetPort) {
     proxyTarget = targetPort
-  if (proxyServer) {
+    console.log(JSON.stringify({ src: 'comfy-host-agent', event: 'proxy-target', listen: proxyPort, target: proxyTarget }))
-    try { proxyServer.close() } catch { /* ignore */ }
-    proxyServer = null
   }
+  if (proxyServer) return
   proxyServer = net.createServer((client) => {
     const upstream = net.connect(proxyTarget, '127.0.0.1')
     const fail = () => {
@@ -85,6 +121,12 @@ function ensureProxy(targetPort) {
   })
 }
 
+async function syncProxy() {
+  const healthyPort = await findHealthyPort()
+  if (healthyPort) ensureProxy(healthyPort)
+  return healthyPort
+}
+
 async function processUp() {
   try {
     const { stdout } = await execFileAsync('tasklist', ['/FO', 'CSV', '/NH'], { windowsHide: true, timeout: 4000 })
@@ -231,8 +273,7 @@ const server = http.createServer(async (req, res) => {
   if (!authorized(req)) return json(res, 401, { ok: false, error: 'unauthorized' })
   const url = new URL(req.url || '/', 'http://localhost')
   if (req.method === 'GET' && url.pathname === '/status') {
-    const [healthyPort, process] = await Promise.all([findHealthyPort(), processUp()])
+    const [healthyPort, process] = await Promise.all([syncProxy(), processUp()])
-    if (healthyPort) ensureProxy(healthyPort)
     return json(res, 200, {
       ok: true,
       http: Boolean(healthyPort),
@@ -242,9 +283,8 @@ const server = http.createServer(async (req, res) => {
     })
   }
   if (req.method === 'POST' && url.pathname === '/start') {
-    const healthyPort = await findHealthyPort()
+    const healthyPort = await syncProxy()
     if (healthyPort) {
-      ensureProxy(healthyPort)
       return json(res, 200, { ok: true, started: false, already: true, port: healthyPort, proxyPort })
     }
     const launched = startComfy()
@@ -258,8 +298,7 @@ const server = http.createServer(async (req, res) => {
 })
 
 server.listen(port, '0.0.0.0', async () => {
-  const healthyPort = await findHealthyPort()
+  const healthyPort = await syncProxy()
-  if (healthyPort) ensureProxy(healthyPort)
   console.log(JSON.stringify({
     src: 'comfy-host-agent',
     event: 'listen',
@@ -268,4 +307,7 @@ server.listen(port, '0.0.0.0', async () => {
     comfyPort: healthyPort || null,
     candidates: candidatePorts()
   }))
+  setInterval(() => {
+    syncProxy().catch(() => {})
+  }, 3000)
 })

scripts/start-comfy-host-agent.cmd

@@ -0,0 +1,3 @@
+@echo off
+cd /d "%~dp0.."
+node "%~dp0comfy-host-agent.mjs"

server/api/v2/generate.post.ts

@@ -0,0 +1,255 @@
+import { addStudioJob, kickStudioQueue, listStudioJobs, videoJobsBusy } from '~/server/utils/studioQueue'
+import { comfyConfigured } from '~/server/utils/comfy'
+import { imageDimensions } from '~/server/utils/resolution'
+import { clampImageCfg, clampImageScaleMegapixels, clampImageSteps } from '~/utils/generationPresets'
+import {
+  IMAGE_V2_CFG_DEFAULT,
+  IMAGE_V2_CONSISTENCY_CLIP,
+  IMAGE_V2_CONSISTENCY_MODEL,
+  IMAGE_V2_SNOFS_CLIP,
+  IMAGE_V2_SNOFS_MODEL,
+  IMAGE_V2_STEPS_DEFAULT,
+  IMAGE_V2_TURBO_CFG,
+  IMAGE_V2_TURBO_STEPS,
+  clampImageV2Strength,
+  parseImageV2Mode,
+  parseImageV2Task,
+  type ImageV2Mode,
+  type ImageV2Task
+} from '~/utils/imageV2'
+import { getStill, rememberInputStill, stillPath } from '~/server/utils/library'
+import { existsSync, readFileSync } from 'node:fs'
+
+type ImageFile = { filename: string; data: Buffer; type?: string }
+
+function parseBool(raw: unknown) {
+  return raw === true || raw === 'true' || raw === '1' || raw === 1
+}
+
+function isHttpUrl(value: string) {
+  return /^https?:\/\//i.test(value)
+}
+
+async function fileFromUrl(url: string): Promise<ImageFile> {
+  const res = await fetch(url, { signal: AbortSignal.timeout(20_000) })
+  if (!res.ok) {
+    throw createError({ statusCode: 400, statusMessage: `Could not fetch image (${res.status})` })
+  }
+  const mime = String(res.headers.get('content-type') || 'image/png').split(';')[0]
+  if (!/^image\//i.test(mime)) {
+    throw createError({ statusCode: 400, statusMessage: 'image_a / image_b URL must be an image' })
+  }
+  const data = Buffer.from(await res.arrayBuffer())
+  if (data.length > 40 * 1024 * 1024) {
+    throw createError({ statusCode: 400, statusMessage: 'Image is too large' })
+  }
+  const ext = mime.includes('jpeg') || mime.includes('jpg') ? '.jpg' : mime.includes('webp') ? '.webp' : '.png'
+  return { filename: `remote${ext}`, data, type: mime }
+}
+
+function fileFromStill(ownerKey: string, id: string): ImageFile {
+  const still = getStill(ownerKey, id)
+  const path = stillPath(ownerKey, still.id)
+  if (!existsSync(path)) {
+    throw createError({ statusCode: 400, statusMessage: 'That still is missing from the library' })
+  }
+  return {
+    filename: still.filename || `${still.id}.png`,
+    data: readFileSync(path),
+    type: 'image/png'
+  }
+}
+
+async function resolveImageRef(ownerKey: string, raw: unknown, uploaded: ImageFile | null) {
+  if (uploaded) return uploaded
+  const value = String(raw || '').trim()
+  if (!value || value === 'null') return null
+  if (isHttpUrl(value)) return fileFromUrl(value)
+  return fileFromStill(ownerKey, value)
+}
+
+function readMultipart(parts: Array<{ name?: string; filename?: string; type?: string; data?: Buffer }> | null) {
+  const fields: Record<string, string> = {}
+  let imageA: ImageFile | null = null
+  let imageB: ImageFile | null = null
+  for (const part of parts || []) {
+    if ((part.name === 'image_a' || part.name === 'image') && part.filename && part.data?.length) {
+      imageA = { filename: part.filename, data: part.data, type: part.type }
+    } else if ((part.name === 'image_b' || part.name === 'image2') && part.filename && part.data?.length) {
+      imageB = { filename: part.filename, data: part.data, type: part.type }
+    } else if (part.name && part.data) {
+      fields[part.name] = part.data.toString('utf8')
+    }
+  }
+  return { fields, imageA, imageB }
+}
+
+export default defineEventHandler(async (event) => {
+  const contentType = String(getHeader(event, 'content-type') || '')
+  let fields: Record<string, unknown> = {}
+  let uploadedA: ImageFile | null = null
+  let uploadedB: ImageFile | null = null
+
+  if (contentType.includes('multipart/form-data')) {
+    const form = await readMultipartFormData(event).catch(() => null)
+    const parsed = readMultipart(form || [])
+    fields = parsed.fields
+    uploadedA = parsed.imageA
+    uploadedB = parsed.imageB
+  } else {
+    fields = await readBody<Record<string, unknown>>(event).catch(() => ({}))
+  }
+
+  const mode = parseImageV2Mode(fields.mode)
+  if (!mode) {
+    throw createError({ statusCode: 400, statusMessage: 'mode must be edit or compose' })
+  }
+  const task = parseImageV2Task(fields.task, 'scene')
+  const prompt = String(fields.prompt || '').trim()
+  if (!prompt) {
+    throw createError({ statusCode: 400, statusMessage: 'A prompt is required' })
+  }
+  if (!comfyConfigured()) {
+    throw createError({
+      statusCode: 503,
+      statusMessage: 'Beast ComfyUI is not configured. Set COMFY_HOST.'
+    })
+  }
+
+  if (mode === 'edit' && (task === 'identity' || task === 'outfit' || task === 'face_lock')) {
+    throw createError({
+      statusCode: 400,
+      statusMessage: `${task} requires Compose and image B. Edit is one image only.`
+    })
+  }
+
+  const ownerKey = libraryOwnerKey(event)
+  const imageA = await resolveImageRef(ownerKey, fields.image_a, uploadedA)
+  const imageB = await resolveImageRef(ownerKey, fields.image_b, uploadedB)
+
+  if (!imageA) {
+    throw createError({ statusCode: 400, statusMessage: 'image_a is required' })
+  }
+  if (mode === 'edit' && imageB) {
+    throw createError({
+      statusCode: 400,
+      statusMessage: 'Edit mode takes one image. Use Compose for two stills.'
+    })
+  }
+  if (mode === 'compose' && !imageB) {
+    throw createError({
+      statusCode: 400,
+      statusMessage: 'Compose requires image_b. Refusing to fall back to one-image generation.'
+    })
+  }
+
+  const library = publicLibrary(event)
+  const folderId = library.folders.some(folder => folder.id === String(fields.folderId || ''))
+    ? String(fields.folderId)
+    : library.folders[0]?.id
+  if (!folderId) {
+    throw createError({ statusCode: 400, statusMessage: 'Create a library folder before generating' })
+  }
+  assertFolderExists(event, folderId)
+
+  const destFolder = library.folders.find(folder => folder.id === folderId)
+  const folderLocked = Boolean(destFolder?.protected && !destFolder.unlocked)
+  const hideInput = parseBool(fields.hideInput)
+  const hideThumbnail = parseBool(fields.hideThumbnail)
+  const turbo = parseBool(fields.turbo)
+  const steps = turbo ? IMAGE_V2_TURBO_STEPS : clampImageSteps(fields.steps, IMAGE_V2_STEPS_DEFAULT)
+  const cfg = turbo ? IMAGE_V2_TURBO_CFG : clampImageCfg(fields.cfg, IMAGE_V2_CFG_DEFAULT)
+  const megapixels = clampImageScaleMegapixels(fields.megapixels ?? fields.scaleMegapixels, 1)
+  const seed = fields.seed && String(fields.seed) !== 'random'
+    ? Number(fields.seed)
+    : Math.floor(Math.random() * 2_147_483_647)
+  const size = imageDimensions(imageA.data)
+  const clipName = String(fields.name || '').trim().slice(0, 80)
+  const v2Mode = mode as ImageV2Mode
+  const v2Task = (mode === 'edit' ? 'scene' : task) as ImageV2Task
+
+  const still = await rememberInputStill({
+    ownerKey,
+    folderId,
+    filename: imageA.filename,
+    data: imageA.data,
+    width: size?.width || 0,
+    height: size?.height || 0,
+    hideInput
+  })
+  const savedRef = imageB
+    ? await rememberInputStill({
+      ownerKey,
+      folderId,
+      filename: imageB.filename,
+      data: imageB.data,
+      hideInput
+    })
+    : null
+
+  const studio = await addStudioJob({
+    ownerKey,
+    familyId: crypto.randomUUID(),
+    kind: 'edit',
+    payload: {
+      prompt,
+      name: clipName,
+      folderId,
+      aspect: 'auto',
+      width: size?.width || 0,
+      height: size?.height || 0,
+      steps,
+      turbo,
+      seed,
+      cfg,
+      fps: 24,
+      samplerName: 'euler',
+      scheduler: 'simple',
+      duration: 0,
+      sound: false,
+      workflow: 'v1',
+      useIdentityRefs: false,
+      stillId: still?.id,
+      stillFilename: still?.filename,
+      hideThumbnail,
+      hideInput,
+      folderLocked,
+      referenceStillIds: [null, null, null, null],
+      extensions: [],
+      queueAutoRun: false,
+      negative: String(fields.negative || '').trim(),
+      referenceStillId: savedRef?.id,
+      referenceStillFilename: savedRef?.filename,
+      scaleToTotalPixels: true,
+      scaleMegapixels: megapixels,
+      imagePipeline: 'v2',
+      v2Mode,
+      v2Task,
+      snofsModel: clampImageV2Strength(fields.snofs_model, IMAGE_V2_SNOFS_MODEL),
+      snofsClip: clampImageV2Strength(fields.snofs_clip, IMAGE_V2_SNOFS_CLIP),
+      consistencyModel: clampImageV2Strength(fields.consistency_model, IMAGE_V2_CONSISTENCY_MODEL),
+      consistencyClip: clampImageV2Strength(fields.consistency_clip, IMAGE_V2_CONSISTENCY_CLIP)
+    }
+  })
+  await kickStudioQueue()
+  let latest = listStudioJobs(ownerKey).find(item => item.id === studio.id)
+  if (!latest?.liveJobId && !(await videoJobsBusy())) {
+    await kickStudioQueue()
+    latest = listStudioJobs(ownerKey).find(item => item.id === studio.id)
+  }
+  const liveJobId = latest?.liveJobId || ''
+
+  return {
+    jobId: liveJobId || studio.id,
+    studioJobId: studio.id,
+    queued: !liveJobId,
+    seed,
+    steps,
+    cfg,
+    mode: v2Mode,
+    task: v2Task,
+    workflow: v2Mode === 'compose' ? 'klein_v2_compose.json' : 'klein_v2_edit.json',
+    hideThumbnail,
+    folderLocked
+  }
+})

server/assets/klein_v2_compose.json

@@ -0,0 +1,210 @@
+{
+  "1": {
+    "inputs": { "image": "" },
+    "class_type": "LoadImage",
+    "_meta": { "title": "Load Image A" }
+  },
+  "22": {
+    "inputs": { "image": "" },
+    "class_type": "LoadImage",
+    "_meta": { "title": "Load Image B" }
+  },
+  "2": {
+    "inputs": {
+      "upscale_method": "lanczos",
+      "megapixels": 1,
+      "resolution_steps": 1,
+      "image": ["1", 0]
+    },
+    "class_type": "ImageScaleToTotalPixels",
+    "_meta": { "title": "Scale Image A" }
+  },
+  "23": {
+    "inputs": {
+      "upscale_method": "lanczos",
+      "megapixels": 1,
+      "resolution_steps": 1,
+      "image": ["22", 0]
+    },
+    "class_type": "ImageScaleToTotalPixels",
+    "_meta": { "title": "Scale Image B" }
+  },
+  "3": {
+    "inputs": { "image": ["2", 0] },
+    "class_type": "GetImageSize",
+    "_meta": { "title": "Get Image Size" }
+  },
+  "4": {
+    "inputs": {
+      "unet_name": "flux-2-klein-base-9b-fp8.safetensors",
+      "weight_dtype": "default"
+    },
+    "class_type": "UNETLoader",
+    "_meta": { "title": "Load Flux.2 Klein 9B Base" }
+  },
+  "5": {
+    "inputs": {
+      "clip_name": "qwen_3_8b_fp8mixed.safetensors",
+      "type": "flux2",
+      "device": "default"
+    },
+    "class_type": "CLIPLoader",
+    "_meta": { "title": "Load Qwen 3 8B CLIP" }
+  },
+  "6": {
+    "inputs": { "vae_name": "full_encoder_small_decoder.safetensors" },
+    "class_type": "VAELoader",
+    "_meta": { "title": "Load Klein VAE" }
+  },
+  "7": {
+    "inputs": {
+      "lora_name": "klein_snofs_v1_4.safetensors",
+      "strength_model": 0.65,
+      "strength_clip": 0.35,
+      "model": ["4", 0],
+      "clip": ["5", 0]
+    },
+    "class_type": "LoraLoader",
+    "_meta": { "title": "SNOFS" }
+  },
+  "8": {
+    "inputs": {
+      "lora_name": "Flux2-Klein-9B-consistency-V2.safetensors",
+      "strength_model": 0.7,
+      "strength_clip": 0.7,
+      "model": ["7", 0],
+      "clip": ["7", 1]
+    },
+    "class_type": "LoraLoader",
+    "_meta": { "title": "Consistency" }
+  },
+  "9": {
+    "inputs": {
+      "text": "",
+      "clip": ["8", 1]
+    },
+    "class_type": "CLIPTextEncode",
+    "_meta": { "title": "Positive Prompt" }
+  },
+  "10": {
+    "inputs": {
+      "text": "",
+      "clip": ["8", 1]
+    },
+    "class_type": "CLIPTextEncode",
+    "_meta": { "title": "Negative Prompt" }
+  },
+  "11": {
+    "inputs": {
+      "pixels": ["2", 0],
+      "vae": ["6", 0]
+    },
+    "class_type": "VAEEncode",
+    "_meta": { "title": "VAE Encode A" }
+  },
+  "24": {
+    "inputs": {
+      "pixels": ["23", 0],
+      "vae": ["6", 0]
+    },
+    "class_type": "VAEEncode",
+    "_meta": { "title": "VAE Encode B" }
+  },
+  "12": {
+    "inputs": {
+      "conditioning": ["9", 0],
+      "latent": ["11", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent A+" }
+  },
+  "25": {
+    "inputs": {
+      "conditioning": ["12", 0],
+      "latent": ["24", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent B+" }
+  },
+  "13": {
+    "inputs": {
+      "conditioning": ["10", 0],
+      "latent": ["11", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent A-" }
+  },
+  "26": {
+    "inputs": {
+      "conditioning": ["13", 0],
+      "latent": ["24", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent B-" }
+  },
+  "14": {
+    "inputs": {
+      "width": ["3", 0],
+      "height": ["3", 1],
+      "batch_size": 1
+    },
+    "class_type": "EmptyFlux2LatentImage",
+    "_meta": { "title": "Empty Flux 2 Latent" }
+  },
+  "15": {
+    "inputs": { "noise_seed": 1 },
+    "class_type": "RandomNoise",
+    "_meta": { "title": "RandomNoise" }
+  },
+  "16": {
+    "inputs": { "sampler_name": "euler" },
+    "class_type": "KSamplerSelect",
+    "_meta": { "title": "KSamplerSelect" }
+  },
+  "17": {
+    "inputs": {
+      "steps": 24,
+      "width": ["3", 0],
+      "height": ["3", 1]
+    },
+    "class_type": "Flux2Scheduler",
+    "_meta": { "title": "Flux2Scheduler" }
+  },
+  "18": {
+    "inputs": {
+      "cfg": 4,
+      "model": ["8", 0],
+      "positive": ["25", 0],
+      "negative": ["26", 0]
+    },
+    "class_type": "CFGGuider",
+    "_meta": { "title": "CFG Guider" }
+  },
+  "19": {
+    "inputs": {
+      "noise": ["15", 0],
+      "guider": ["18", 0],
+      "sampler": ["16", 0],
+      "sigmas": ["17", 0],
+      "latent_image": ["14", 0]
+    },
+    "class_type": "SamplerCustomAdvanced",
+    "_meta": { "title": "SamplerCustomAdvanced" }
+  },
+  "20": {
+    "inputs": {
+      "samples": ["19", 0],
+      "vae": ["6", 0]
+    },
+    "class_type": "VAEDecode",
+    "_meta": { "title": "VAE Decode" }
+  },
+  "21": {
+    "inputs": {
+      "filename_prefix": "aigen-v2",
+      "images": ["20", 0]
+    },
+    "class_type": "SaveImage",
+    "_meta": { "title": "Save Image" }
+  }
+}

server/assets/klein_v2_edit.json

@@ -0,0 +1,171 @@
+{
+  "1": {
+    "inputs": { "image": "" },
+    "class_type": "LoadImage",
+    "_meta": { "title": "Load Image A" }
+  },
+  "2": {
+    "inputs": {
+      "upscale_method": "lanczos",
+      "megapixels": 1,
+      "resolution_steps": 1,
+      "image": ["1", 0]
+    },
+    "class_type": "ImageScaleToTotalPixels",
+    "_meta": { "title": "Scale Image A" }
+  },
+  "3": {
+    "inputs": { "image": ["2", 0] },
+    "class_type": "GetImageSize",
+    "_meta": { "title": "Get Image Size" }
+  },
+  "4": {
+    "inputs": {
+      "unet_name": "flux-2-klein-base-9b-fp8.safetensors",
+      "weight_dtype": "default"
+    },
+    "class_type": "UNETLoader",
+    "_meta": { "title": "Load Flux.2 Klein 9B Base" }
+  },
+  "5": {
+    "inputs": {
+      "clip_name": "qwen_3_8b_fp8mixed.safetensors",
+      "type": "flux2",
+      "device": "default"
+    },
+    "class_type": "CLIPLoader",
+    "_meta": { "title": "Load Qwen 3 8B CLIP" }
+  },
+  "6": {
+    "inputs": { "vae_name": "full_encoder_small_decoder.safetensors" },
+    "class_type": "VAELoader",
+    "_meta": { "title": "Load Klein VAE" }
+  },
+  "7": {
+    "inputs": {
+      "lora_name": "klein_snofs_v1_4.safetensors",
+      "strength_model": 0.65,
+      "strength_clip": 0.35,
+      "model": ["4", 0],
+      "clip": ["5", 0]
+    },
+    "class_type": "LoraLoader",
+    "_meta": { "title": "SNOFS" }
+  },
+  "8": {
+    "inputs": {
+      "lora_name": "Flux2-Klein-9B-consistency-V2.safetensors",
+      "strength_model": 0.7,
+      "strength_clip": 0.7,
+      "model": ["7", 0],
+      "clip": ["7", 1]
+    },
+    "class_type": "LoraLoader",
+    "_meta": { "title": "Consistency" }
+  },
+  "9": {
+    "inputs": {
+      "text": "",
+      "clip": ["8", 1]
+    },
+    "class_type": "CLIPTextEncode",
+    "_meta": { "title": "Positive Prompt" }
+  },
+  "10": {
+    "inputs": {
+      "text": "",
+      "clip": ["8", 1]
+    },
+    "class_type": "CLIPTextEncode",
+    "_meta": { "title": "Negative Prompt" }
+  },
+  "11": {
+    "inputs": {
+      "pixels": ["2", 0],
+      "vae": ["6", 0]
+    },
+    "class_type": "VAEEncode",
+    "_meta": { "title": "VAE Encode A" }
+  },
+  "12": {
+    "inputs": {
+      "conditioning": ["9", 0],
+      "latent": ["11", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent A+" }
+  },
+  "13": {
+    "inputs": {
+      "conditioning": ["10", 0],
+      "latent": ["11", 0]
+    },
+    "class_type": "ReferenceLatent",
+    "_meta": { "title": "Reference Latent A-" }
+  },
+  "14": {
+    "inputs": {
+      "width": ["3", 0],
+      "height": ["3", 1],
+      "batch_size": 1
+    },
+    "class_type": "EmptyFlux2LatentImage",
+    "_meta": { "title": "Empty Flux 2 Latent" }
+  },
+  "15": {
+    "inputs": { "noise_seed": 1 },
+    "class_type": "RandomNoise",
+    "_meta": { "title": "RandomNoise" }
+  },
+  "16": {
+    "inputs": { "sampler_name": "euler" },
+    "class_type": "KSamplerSelect",
+    "_meta": { "title": "KSamplerSelect" }
+  },
+  "17": {
+    "inputs": {
+      "steps": 24,
+      "width": ["3", 0],
+      "height": ["3", 1]
+    },
+    "class_type": "Flux2Scheduler",
+    "_meta": { "title": "Flux2Scheduler" }
+  },
+  "18": {
+    "inputs": {
+      "cfg": 4,
+      "model": ["8", 0],
+      "positive": ["12", 0],
+      "negative": ["13", 0]
+    },
+    "class_type": "CFGGuider",
+    "_meta": { "title": "CFG Guider" }
+  },
+  "19": {
+    "inputs": {
+      "noise": ["15", 0],
+      "guider": ["18", 0],
+      "sampler": ["16", 0],
+      "sigmas": ["17", 0],
+      "latent_image": ["14", 0]
+    },
+    "class_type": "SamplerCustomAdvanced",
+    "_meta": { "title": "SamplerCustomAdvanced" }
+  },
+  "20": {
+    "inputs": {
+      "samples": ["19", 0],
+      "vae": ["6", 0]
+    },
+    "class_type": "VAEDecode",
+    "_meta": { "title": "VAE Decode" }
+  },
+  "21": {
+    "inputs": {
+      "filename_prefix": "aigen-v2",
+      "images": ["20", 0]
+    },
+    "class_type": "SaveImage",
+    "_meta": { "title": "Save Image" }
+  }
+}

server/routes/v2/generate.post.ts

@@ -0,0 +1 @@
+export { default } from '../../api/v2/generate.post'

server/utils/comfy.ts

@@ -15,8 +15,8 @@ export function comfyConfigured() {
 function comfyBase() {
   if (comfyHostOverride) return comfyHostOverride
   const config = useRuntimeConfig()
-  let host = String(config.comfyHost || process.env.COMFY_HOST || '').trim().replace(/\/$/, '')
+  let host = String(process.env.COMFY_HOST || config.comfyHost || '').trim().replace(/\/$/, '')
-  const port = String(config.comfyPort || process.env.COMFY_PORT || '').trim()
+  const port = String(process.env.COMFY_PORT || config.comfyPort || '').trim()
   if (!host) {
     throw createError({ statusCode: 500, statusMessage: 'COMFY_HOST is not configured' })
   }

server/utils/generationPresets.ts

@@ -14,6 +14,18 @@ import {
   type ImagePresetSettings,
   type VideoPresetSettings
 } from '~/utils/generationPresets'
+import {
+  IMAGE_V2_CFG_DEFAULT,
+  IMAGE_V2_CONSISTENCY_CLIP,
+  IMAGE_V2_CONSISTENCY_MODEL,
+  IMAGE_V2_SNOFS_CLIP,
+  IMAGE_V2_SNOFS_MODEL,
+  IMAGE_V2_STEPS_DEFAULT,
+  clampImageV2Strength,
+  parseImageV2Mode,
+  parseImageV2Task,
+  type ImageV2PresetSettings
+} from '~/utils/imageV2'
 import { coerceVideoWorkflow, isXaigenStudio, parseVideoWorkflow } from '~/utils/videoModels'
 
 const writeChains = new Map<string, Promise<unknown>>()
@@ -96,6 +108,31 @@ function sanitizeImageSettings(raw: unknown): ImagePresetSettings {
   }
 }
 
+function sanitizeImageV2Settings(raw: unknown): ImageV2PresetSettings {
+  const rec = raw && typeof raw === 'object' ? raw as Record<string, unknown> : {}
+  const mode = parseImageV2Mode(rec.mode) || 'edit'
+  const turbo = rec.turbo === true
+  return {
+    mode,
+    task: mode === 'compose' ? parseImageV2Task(rec.task, 'scene') : 'scene',
+    negative: String(rec.negative || '').slice(0, 2000),
+    snofsModel: clampImageV2Strength(rec.snofsModel ?? rec.snofs_model, IMAGE_V2_SNOFS_MODEL),
+    snofsClip: clampImageV2Strength(rec.snofsClip ?? rec.snofs_clip, IMAGE_V2_SNOFS_CLIP),
+    consistencyModel: clampImageV2Strength(rec.consistencyModel ?? rec.consistency_model, IMAGE_V2_CONSISTENCY_MODEL),
+    consistencyClip: clampImageV2Strength(rec.consistencyClip ?? rec.consistency_clip, IMAGE_V2_CONSISTENCY_CLIP),
+    steps: turbo ? 8 : clampImageSteps(rec.steps, IMAGE_V2_STEPS_DEFAULT),
+    cfg: turbo ? 1 : clampImageCfg(rec.cfg, IMAGE_V2_CFG_DEFAULT),
+    megapixels: clampImageScaleMegapixels(rec.megapixels ?? rec.scaleMegapixels, 1),
+    turbo
+  }
+}
+
+function sanitizeSettings(kind: GenerationPresetKind, raw: unknown) {
+  if (kind === 'image') return sanitizeImageSettings(raw)
+  if (kind === 'imagev2') return sanitizeImageV2Settings(raw)
+  return sanitizeVideoSettings(raw)
+}
+
 function normalizeStoredPreset(raw: unknown): GenerationPreset | null {
   if (!raw || typeof raw !== 'object') return null
   const rec = raw as Record<string, unknown>
@@ -111,7 +148,7 @@ function normalizeStoredPreset(raw: unknown): GenerationPreset | null {
     createdAt,
     updatedAt: Number(rec.updatedAt) || createdAt,
     loraStack: normalizeLoraStack(rec.loraStack),
-    settings: kind === 'image' ? sanitizeImageSettings(rec.settings) : sanitizeVideoSettings(rec.settings)
+    settings: sanitizeSettings(kind, rec.settings)
   }
 }
 
@@ -147,7 +184,7 @@ export async function saveGenerationPreset(owner: string, input: {
   const kind = parseGenerationPresetKind(input.kind)
   const xaigen = isXaigenStudio()
   const loraStack = filterLoraStackForStudio(input.loraStack, xaigen)
-  const settings = kind === 'image' ? sanitizeImageSettings(input.settings) : sanitizeVideoSettings(input.settings)
+  const settings = sanitizeSettings(kind, input.settings)
   const now = Date.now()
   return mutate(owner, (presets) => {
     const existing = presets.find(item => item.kind === kind && item.name.toLowerCase() === name.toLowerCase())

server/utils/imageChainV2.ts

@@ -0,0 +1,168 @@
+import { createJob, emitJob, type Job } from '~/server/utils/jobs'
+import { ensureComfyReady } from '~/server/utils/comfyLifecycle'
+import { assertImageScaleToTotalPixelsNode, getComfyHost, uploadImage, queuePrompt, purgeComfyArtifacts } from '~/server/utils/comfy'
+import { withImageComfyHost, waitForImageEdit, downloadEditedImage } from '~/server/utils/imageComfy'
+import { buildImageV2Workflow } from '~/server/utils/imageWorkflowV2'
+import { ensureComfyLoraNames } from '~/server/utils/loras'
+import { imageDimensions } from '~/server/utils/resolution'
+import { emitChainJob } from '~/server/utils/watch'
+import { saveStill } from '~/server/utils/library'
+import type { ImageV2Mode, ImageV2Task } from '~/utils/imageV2'
+import type { EditImageFile } from '~/server/utils/imageChain'
+
+export type EditV2RunParams = {
+  mode: ImageV2Mode
+  task: ImageV2Task
+  image: EditImageFile
+  reference: EditImageFile | null
+  prompt: string
+  negative: string
+  steps: number
+  seed: number
+  cfg: number
+  snofsModel: number
+  snofsClip: number
+  consistencyModel: number
+  consistencyClip: number
+  megapixels: number
+}
+
+export async function runEditV2(job: Job, params: EditV2RunParams) {
+  const library = job.library
+  if (!library) throw new Error('Edit job is missing library metadata')
+  if (params.mode === 'compose' && !params.reference) {
+    throw new Error('Compose requires image B. Refusing to fall back to one-image generation.')
+  }
+  if (params.mode === 'edit' && params.reference) {
+    throw new Error('Edit mode takes one image. Use Compose for two stills.')
+  }
+
+  try {
+    await ensureComfyReady((status) => {
+      emitChainJob(job, {
+        type: status.state === 'busy' ? 'busy' : 'status',
+        message: status.message,
+        progress: status.state === 'online' ? Math.max(job.progress, 6) : Math.max(job.progress, 3),
+        busy: status.state === 'busy'
+      })
+    }, { skipBusyWait: true })
+    if (job.status === 'cancelled') throw new Error('Job interrupted.')
+    job.imageComfyHost = getComfyHost()
+
+    await withImageComfyHost(job.imageComfyHost, async () => {
+      emitChainJob(job, {
+        type: 'status',
+        message: params.mode === 'compose' ? 'Uploading stills A and B to Beast...' : 'Uploading still A to Beast...',
+        progress: 8
+      })
+      const uploaded = await uploadImage(params.image, job.id)
+      const uploadedRef = params.reference
+        ? await uploadImage({
+          ...params.reference,
+          filename: `ref_${params.reference.filename || 'image_b.png'}`
+        }, job.id)
+        : null
+      if (job.status === 'cancelled') throw new Error('Job interrupted.')
+
+      emitChainJob(job, {
+        type: 'status',
+        message: params.mode === 'compose'
+          ? `Queueing Klein v2 compose (${params.task}) on Beast...`
+          : 'Queueing Klein v2 edit on Beast...',
+        progress: 12
+      })
+      await ensureComfyLoraNames('image')
+      await assertImageScaleToTotalPixelsNode()
+      const built = buildImageV2Workflow({
+        mode: params.mode,
+        task: params.task,
+        prompt: params.prompt,
+        negative: params.negative,
+        imageAName: uploaded.name,
+        imageBName: uploadedRef?.name,
+        snofsModel: params.snofsModel,
+        snofsClip: params.snofsClip,
+        consistencyModel: params.consistencyModel,
+        consistencyClip: params.consistencyClip,
+        steps: params.steps,
+        cfg: params.cfg,
+        seed: params.seed,
+        megapixels: params.megapixels,
+        filenamePrefix: `aigen_v2_${job.id.slice(0, 8)}`
+      })
+      const queued = await queuePrompt(built.graph, job.clientId)
+      job.promptId = queued.prompt_id
+      job.status = 'running'
+      emitChainJob(job, {
+        type: 'status',
+        message: `Running ${built.workflowFile}...`,
+        progress: 18,
+        maxStep: params.steps
+      })
+
+      const output = await waitForImageEdit({
+        promptId: queued.prompt_id,
+        clientId: job.clientId,
+        timeoutMs: 10 * 60 * 1000,
+        onProgress: (event) => {
+          emitChainJob(job, {
+            type: 'status',
+            message: event.message,
+            progress: event.progress,
+            step: event.step,
+            maxStep: event.maxStep || params.steps,
+            node: event.node
+          })
+        },
+        isCancelled: () => job.status === 'cancelled'
+      })
+
+      emitChainJob(job, { type: 'status', message: 'Saving Klein v2 still...', progress: 94 })
+      const buffer = await downloadEditedImage(output)
+      const size = imageDimensions(buffer)
+      const still = await saveStill({
+        ownerKey: library.ownerKey,
+        folderId: library.folderId,
+        filename: output.filename,
+        data: buffer,
+        width: size?.width || 0,
+        height: size?.height || 0,
+        hideInput: job.hideThumbnail === true,
+        role: 'output',
+        name: library.name || undefined,
+        prompt: built.prompt
+      })
+      job.stillId = still?.id
+      await purgeComfyArtifacts({
+        video: { filename: output.filename, subfolder: output.subfolder, type: output.type },
+        imageName: uploaded.name,
+        imageSubfolder: uploaded.subfolder,
+        extraImageNames: uploadedRef?.name ? [uploadedRef.name] : [],
+        promptId: job.promptId
+      })
+
+      job.status = 'complete'
+      emitChainJob(job, {
+        type: 'complete',
+        message: 'Klein v2 finished on Beast',
+        progress: 100,
+        stillId: still?.id,
+        filename: output.filename,
+        subfolder: output.subfolder,
+        mediaType: 'image',
+        hideThumbnail: job.hideThumbnail,
+        folderLocked: library.folderLocked
+      })
+    })
+  } catch (error) {
+    if (job.status !== 'error' && job.status !== 'cancelled') {
+      const message = error instanceof Error ? error.message : String(error)
+      job.status = 'error'
+      job.error = message
+      emitJob(job, { type: 'error', error: message, message })
+    }
+  } finally {
+    const { onLiveVideoSettled } = await import('~/server/utils/studioQueue')
+    await onLiveVideoSettled(job)
+  }
+}

server/utils/imageComfy.ts

@@ -292,7 +292,7 @@ export async function waitForImageEdit(opts: {
       if (payload.type === 'executing' && payload.data?.node) {
         opts.onProgress?.({
           message: 'Running Flux.2 Klein...',
-          progress: payload.data.node === '9' || payload.data.node === '94' ? 92 : 30,
+          progress: payload.data.node === '9' || payload.data.node === '94' || payload.data.node === '21' ? 92 : 30,
           node: payload.data.node
         })
       }

server/utils/imageWorkflowV2.ts

@@ -0,0 +1,193 @@
+import editTemplate from '../assets/klein_v2_edit.json'
+import composeTemplate from '../assets/klein_v2_compose.json'
+import { IMAGE_SCALE_TO_TOTAL_PIXELS } from '~/server/utils/comfy'
+import { cachedComfyLoraNames } from '~/server/utils/loras'
+import { resolveComfyLoraName, loraIdentityKey } from '~/utils/loras'
+import { clampImageCfg, clampImageScaleMegapixels, clampImageSteps } from '~/utils/generationPresets'
+import {
+  IMAGE_V2_CONSISTENCY_CLIP,
+  IMAGE_V2_CONSISTENCY_LORA,
+  IMAGE_V2_CONSISTENCY_MODEL,
+  IMAGE_V2_SNOFS_CLIP,
+  IMAGE_V2_SNOFS_LORA,
+  IMAGE_V2_SNOFS_MODEL,
+  clampImageV2Strength,
+  composeImageV2Prompt,
+  type ImageV2Mode,
+  type ImageV2Task
+} from '~/utils/imageV2'
+
+type WorkflowNode = { class_type: string; inputs: Record<string, unknown>; _meta?: { title?: string } }
+type WorkflowGraph = Record<string, WorkflowNode>
+
+const LOAD_A = '1'
+const LOAD_B = '22'
+const SCALE_A = '2'
+const SCALE_B = '23'
+const PROMPT = '9'
+const NEGATIVE = '10'
+const NOISE = '15'
+const SCHEDULER = '17'
+const CFG = '18'
+const SAVE = '21'
+const SNOFS = '7'
+const CONSISTENCY = '8'
+
+export const IMAGE_V2_EDIT_WORKFLOW = 'klein_v2_edit.json'
+export const IMAGE_V2_COMPOSE_WORKFLOW = 'klein_v2_compose.json'
+
+export interface ImageV2BuildParams {
+  mode: ImageV2Mode
+  task: ImageV2Task
+  prompt: string
+  negative?: string
+  imageAName: string
+  imageBName?: string
+  snofsModel?: number
+  snofsClip?: number
+  consistencyModel?: number
+  consistencyClip?: number
+  steps: number
+  cfg: number
+  seed: number
+  megapixels?: number
+  filenamePrefix?: string
+}
+
+function setInput(graph: WorkflowGraph, id: string, key: string, value: unknown) {
+  if (graph[id]) graph[id].inputs[key] = value
+}
+
+function loadImageNames(graph: WorkflowGraph) {
+  return Object.entries(graph)
+    .filter(([, node]) => node.class_type === 'LoadImage')
+    .map(([id, node]) => ({
+      id,
+      title: String(node._meta?.title || id),
+      image: String(node.inputs.image || '')
+    }))
+}
+
+function resolveRequiredLora(wanted: string, label: string) {
+  const names = cachedComfyLoraNames('image')
+  const resolved = names.length ? resolveComfyLoraName(wanted, names) : wanted
+  const hit = names.some(name => loraIdentityKey(name) === loraIdentityKey(wanted) || loraIdentityKey(name) === loraIdentityKey(resolved))
+  if (names.length && !hit) {
+    throw createError({ statusCode: 503, statusMessage: `Missing ${label} LoRA (${wanted}) on Beast Comfy` })
+  }
+  return resolved
+}
+
+function patchScaleMegapixels(graph: WorkflowGraph, megapixels: number) {
+  const mp = clampImageScaleMegapixels(megapixels)
+  for (const node of Object.values(graph)) {
+    if (node.class_type !== IMAGE_SCALE_TO_TOTAL_PIXELS) continue
+    node.inputs.megapixels = mp
+    node.inputs.upscale_method = 'lanczos'
+  }
+}
+
+export function assertImageV2Graph(graph: WorkflowGraph, mode: ImageV2Mode, imageBName?: string) {
+  const loaders = loadImageNames(graph)
+  if (mode === 'compose') {
+    if (loaders.length < 2) {
+      throw createError({
+        statusCode: 500,
+        statusMessage: 'Compose graph has no second image input. Refusing to run a one-image fallback.'
+      })
+    }
+    const b = graph[LOAD_B]
+    if (!b || b.class_type !== 'LoadImage' || !String(b.inputs.image || '').trim()) {
+      throw createError({
+        statusCode: 500,
+        statusMessage: 'Compose job is missing Load Image B. Refusing to run.'
+      })
+    }
+  }
+  if (imageBName && loaders.length < 2) {
+    throw createError({
+      statusCode: 500,
+      statusMessage: 'image_b was sent but the executed graph has no second image input.'
+    })
+  }
+  const promptNode = graph[PROMPT]
+  if (!promptNode || promptNode.class_type !== 'CLIPTextEncode') {
+    throw createError({ statusCode: 500, statusMessage: 'v2 graph is missing the positive CLIPTextEncode node.' })
+  }
+  if (Array.isArray(promptNode.inputs.text)) {
+    throw createError({ statusCode: 500, statusMessage: 'v2 prompt is a subgraph link. Refusing to run with a leftover widget prompt.' })
+  }
+  if (!String(promptNode.inputs.text || '').trim()) {
+    throw createError({ statusCode: 400, statusMessage: 'v2 prompt was not patched onto the graph.' })
+  }
+}
+
+export function buildImageV2Workflow(params: ImageV2BuildParams) {
+  const compose = params.mode === 'compose'
+  const graph = structuredClone(compose ? composeTemplate : editTemplate) as WorkflowGraph
+  const prompt = composeImageV2Prompt(params.mode, params.task, params.prompt)
+  const negative = String(params.negative || '')
+  const snofsModel = clampImageV2Strength(params.snofsModel, IMAGE_V2_SNOFS_MODEL)
+  const snofsClip = clampImageV2Strength(params.snofsClip, IMAGE_V2_SNOFS_CLIP)
+  const consistencyModel = clampImageV2Strength(params.consistencyModel, IMAGE_V2_CONSISTENCY_MODEL)
+  const consistencyClip = clampImageV2Strength(params.consistencyClip, IMAGE_V2_CONSISTENCY_CLIP)
+  const steps = clampImageSteps(params.steps, 24)
+  const cfg = clampImageCfg(params.cfg, 4)
+  const workflowFile = compose ? IMAGE_V2_COMPOSE_WORKFLOW : IMAGE_V2_EDIT_WORKFLOW
+
+  setInput(graph, LOAD_A, 'image', params.imageAName)
+  if (compose) setInput(graph, LOAD_B, 'image', params.imageBName || '')
+  setInput(graph, PROMPT, 'text', prompt)
+  setInput(graph, NEGATIVE, 'text', negative)
+  setInput(graph, NOISE, 'noise_seed', params.seed)
+  setInput(graph, SCHEDULER, 'steps', steps)
+  setInput(graph, CFG, 'cfg', cfg)
+  setInput(graph, SAVE, 'filename_prefix', params.filenamePrefix || 'aigen-v2')
+  patchScaleMegapixels(graph, params.megapixels ?? 1)
+
+  setInput(graph, SNOFS, 'lora_name', resolveRequiredLora(IMAGE_V2_SNOFS_LORA, 'SNOFS'))
+  setInput(graph, SNOFS, 'strength_model', snofsModel)
+  setInput(graph, SNOFS, 'strength_clip', snofsClip)
+  setInput(graph, CONSISTENCY, 'lora_name', resolveRequiredLora(IMAGE_V2_CONSISTENCY_LORA, 'Consistency'))
+  setInput(graph, CONSISTENCY, 'strength_model', consistencyModel)
+  setInput(graph, CONSISTENCY, 'strength_clip', consistencyClip)
+
+  assertImageV2Graph(graph, params.mode, params.imageBName)
+
+  const loaders = loadImageNames(graph)
+  console.log(JSON.stringify({
+    src: 'image-v2',
+    workflow: workflowFile,
+    mode: params.mode,
+    task: params.task,
+    loadImage: Object.fromEntries(loaders.map(item => [item.id, { title: item.title, file: item.image }])),
+    loras: {
+      snofs: { name: graph[SNOFS]?.inputs.lora_name, model: snofsModel, clip: snofsClip },
+      consistency: { name: graph[CONSISTENCY]?.inputs.lora_name, model: consistencyModel, clip: consistencyClip }
+    },
+    steps,
+    cfg,
+    seed: params.seed,
+    megapixels: clampImageScaleMegapixels(params.megapixels ?? 1)
+  }))
+
+  return { graph, workflowFile, loaders, prompt }
+}
+
+export const IMAGE_V2_NODE_LABELS: Record<string, string> = {
+  '1': 'Loading image A',
+  '22': 'Loading image B',
+  '2': 'Scaling image A',
+  '23': 'Scaling image B',
+  '4': 'Loading Flux.2 Klein 9B',
+  '5': 'Loading CLIP',
+  '6': 'Loading VAE',
+  '7': 'Applying SNOFS',
+  '8': 'Applying Consistency',
+  '9': 'Encoding prompt',
+  '11': 'Encoding image A',
+  '24': 'Encoding image B',
+  '19': 'Sampling Klein v2',
+  '20': 'Decoding still',
+  '21': 'Saving still'
+}

server/utils/studioQueue.ts

@@ -51,6 +51,13 @@ export interface StudioJobPayload {
   referenceStillFilename?: string
   scaleToTotalPixels?: boolean
   scaleMegapixels?: number
+  imagePipeline?: 'v1' | 'v2'
+  v2Mode?: 'edit' | 'compose'
+  v2Task?: 'scene' | 'identity' | 'outfit' | 'face_lock'
+  snofsModel?: number
+  snofsClip?: number
+  consistencyModel?: number
+  consistencyClip?: number
 }
 
 export interface StudioJob {
@@ -210,6 +217,7 @@ export function summarizeStudioJob(job: StudioJob) {
     liveJobId: job.liveJobId,
     stillId: job.payload.stillId,
     workflow: job.payload.workflow,
+    imagePipeline: job.payload.imagePipeline || 'v1',
     duration: job.payload.duration,
     hideThumbnail: job.payload.hideThumbnail,
     folderLocked: job.payload.folderLocked === true,
@@ -665,6 +673,7 @@ async function startStudioEditJob(item: StudioJob) {
     const { stillPath } = await import('~/server/utils/library')
     const { existsSync, readFileSync } = await import('node:fs')
     const { createEditLiveJob, runEdit } = await import('~/server/utils/imageChain')
+    const { runEditV2 } = await import('~/server/utils/imageChainV2')
 
     const payload = item.payload
     if (!payload.stillId || !existsSync(stillPath(item.ownerKey, payload.stillId))) {
@@ -684,6 +693,66 @@ async function startStudioEditJob(item: StudioJob) {
       }
       : null
 
+    if (payload.imagePipeline === 'v2') {
+      const mode = payload.v2Mode === 'compose' ? 'compose' : 'edit'
+      if (mode === 'compose' && !reference) {
+        throw new Error('Compose requires image B. Refusing to fall back to one-image generation.')
+      }
+      if (mode === 'edit' && reference) {
+        throw new Error('Edit mode takes one image. Use Compose for two stills.')
+      }
+      live = createEditLiveJob({
+        steps: payload.steps,
+        hideThumbnail: payload.hideThumbnail,
+        library: {
+          ownerKey: item.ownerKey,
+          folderId: payload.folderId,
+          hideThumbnail: payload.hideThumbnail,
+          hideInput: payload.hideInput,
+          folderLocked: payload.folderLocked,
+          name: payload.name,
+          prompt: payload.prompt,
+          aspect: payload.aspect || 'auto',
+          width: payload.width,
+          height: payload.height,
+          steps: payload.steps,
+          turbo: payload.turbo === true,
+          seed: payload.seed || Math.floor(Math.random() * 2_147_483_647),
+          cfg: payload.cfg,
+          stillId: payload.stillId,
+          stillFilename: payload.stillFilename,
+          familyId: item.familyId,
+          chainIndex: 0,
+          chainStep: 1,
+          chainTotal: 1
+        }
+      })
+      await markStudioLive(item.ownerKey, item.id, live.id)
+      void runEditV2(live, {
+        mode,
+        task: payload.v2Task || 'scene',
+        image,
+        reference,
+        prompt: payload.prompt,
+        negative: payload.negative || '',
+        steps: payload.steps,
+        seed: live.library?.seed || payload.seed,
+        cfg: payload.cfg,
+        snofsModel: payload.snofsModel ?? 0.65,
+        snofsClip: payload.snofsClip ?? 0.35,
+        consistencyModel: payload.consistencyModel ?? 0.7,
+        consistencyClip: payload.consistencyClip ?? 0.7,
+        megapixels: payload.scaleMegapixels ?? 1
+      }).catch((error) => {
+        const message = error instanceof Error ? error.message : String(error)
+        if (live && live.status !== 'error' && live.status !== 'cancelled' && live.status !== 'deferred') {
+          live.status = 'error'
+          live.error = message
+        }
+      })
+      return
+    }
+
     const passes = payload.passes || []
     live = createEditLiveJob({
       steps: payload.steps,

utils/generationPresets.ts

@@ -1,7 +1,8 @@
 import type { LoraStackItem } from '~/utils/loras'
 import type { VideoWorkflowId } from '~/utils/videoModels'
+import type { ImageV2PresetSettings } from '~/utils/imageV2'
 
-export type GenerationPresetKind = 'video' | 'image'
+export type GenerationPresetKind = 'video' | 'image' | 'imagev2'
 
 export type VideoPresetSettings = {
   workflow?: VideoWorkflowId
@@ -75,11 +76,13 @@ export type GenerationPreset = {
   createdAt: number
   updatedAt: number
   loraStack: LoraStackItem[]
-  settings: VideoPresetSettings | ImagePresetSettings
+  settings: VideoPresetSettings | ImagePresetSettings | ImageV2PresetSettings
 }
 
 export function parseGenerationPresetKind(raw: unknown): GenerationPresetKind {
-  return raw === 'image' ? 'image' : 'video'
+  if (raw === 'image') return 'image'
+  if (raw === 'imagev2') return 'imagev2'
+  return 'video'
 }
 
 export function normalizePresetName(raw: unknown) {

utils/imageV2.ts

@@ -0,0 +1,64 @@
+export const IMAGE_V2_MODES = ['edit', 'compose'] as const
+export const IMAGE_V2_TASKS = ['scene', 'identity', 'outfit', 'face_lock'] as const
+
+export type ImageV2Mode = (typeof IMAGE_V2_MODES)[number]
+export type ImageV2Task = (typeof IMAGE_V2_TASKS)[number]
+
+export const IMAGE_V2_STEPS_DEFAULT = 24
+export const IMAGE_V2_CFG_DEFAULT = 4
+export const IMAGE_V2_TURBO_STEPS = 8
+export const IMAGE_V2_TURBO_CFG = 1
+export const IMAGE_V2_SNOFS_MODEL = 0.65
+export const IMAGE_V2_SNOFS_CLIP = 0.35
+export const IMAGE_V2_CONSISTENCY_MODEL = 0.7
+export const IMAGE_V2_CONSISTENCY_CLIP = 0.7
+export const IMAGE_V2_STRENGTH_MIN = 0
+export const IMAGE_V2_STRENGTH_MAX = 2
+export const IMAGE_V2_STRENGTH_STEP = 0.05
+
+export const IMAGE_V2_SNOFS_LORA = 'klein_snofs_v1_4.safetensors'
+export const IMAGE_V2_CONSISTENCY_LORA = 'Flux2-Klein-9B-consistency-V2.safetensors'
+
+export const IMAGE_V2_ROLE_HEADERS: Record<Exclude<ImageV2Task, 'scene'>, string> = {
+  outfit: 'Person, face, body, pose, and background from image 1. Clothing only from image 2. Fit the outfit from image 2 to the body in image 1. Do not copy image 2’s face, body shape, or pose.',
+  face_lock: 'Body, pose, and scene from image 1. Exact face from image 2.',
+  identity: 'Same person as image 1. Use image 2 only to reinforce the face. Follow the user’s pose/scene prompt.'
+}
+
+export function parseImageV2Mode(raw: unknown): ImageV2Mode | null {
+  const value = String(raw || '').trim().toLowerCase()
+  return IMAGE_V2_MODES.includes(value as ImageV2Mode) ? value as ImageV2Mode : null
+}
+
+export function parseImageV2Task(raw: unknown, fallback: ImageV2Task = 'scene'): ImageV2Task {
+  const value = String(raw || '').trim().toLowerCase()
+  return IMAGE_V2_TASKS.includes(value as ImageV2Task) ? value as ImageV2Task : fallback
+}
+
+export function clampImageV2Strength(raw: unknown, fallback: number) {
+  const value = Number(raw)
+  if (!Number.isFinite(value)) return fallback
+  const snapped = Math.round(value / IMAGE_V2_STRENGTH_STEP) * IMAGE_V2_STRENGTH_STEP
+  return Math.min(IMAGE_V2_STRENGTH_MAX, Math.max(IMAGE_V2_STRENGTH_MIN, Math.round(snapped * 100) / 100))
+}
+
+export function composeImageV2Prompt(mode: ImageV2Mode, task: ImageV2Task, prompt: string) {
+  const body = String(prompt || '').trim()
+  if (mode !== 'compose' || task === 'scene') return body
+  const header = IMAGE_V2_ROLE_HEADERS[task]
+  return header ? `${header}\n\n${body}` : body
+}
+
+export type ImageV2PresetSettings = {
+  mode?: ImageV2Mode
+  task?: ImageV2Task
+  negative?: string
+  snofsModel?: number
+  snofsClip?: number
+  consistencyModel?: number
+  consistencyClip?: number
+  steps?: number
+  cfg?: number
+  megapixels?: number
+  turbo?: boolean
+}