Add Image v2 Generate for Klein text-to-image.

New sibling graph and mode so T2I does not borrow Edit, Compose, or Refine. Stills are ignored; size is width x height only.

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

docs/image-v2.md

@@ -6,10 +6,11 @@ Image edit (v1) and Image v2 are siblings. v1 is unchanged. Do not mute-fix `wor
 |---|---|---|
 | 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`, `klein_v2_compose.json`, or `klein_v2_refine.json` |
+| Graphs | One file, prune unused branch | `klein_v2_edit.json`, `klein_v2_compose.json`, `klein_v2_refine.json`, or `klein_v2_generate.json` |
 | Second still | Optional; switches branch | Compose only; required. Mismatch is a 400 |
 | Mask | — | Refine only; required. Missing mask is a 400, not an Edit fallback |
-| LoRAs | Optional user picker | SNOFS + Consistency, four strengths |
+| Input still | Required | Edit / Compose / Refine required. Generate ignores stills |
+| LoRAs | Optional user picker | SNOFS + Consistency, four strengths. Generate defaults Consistency off |
 | 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}`.
@@ -19,6 +20,7 @@ Poll v2 jobs the same way as v1: `GET /api/generate/{id}/stream` and `GET /api/g
 - **edit** — one image + text. Task is `scene`. Sending `image_b` is rejected. Use this to change lighting, background, or the whole frame from a single still.
 - **compose** — two images + text. `image_b` is required. No silent one-image fallback. Use this to lock a person from A and pull a face or outfit from B.
 - **refine** — one canvas + a painted mask + text. `image_a` and `mask` are required. `image_b` is ignored. Prompt only what should change inside the mask. Strength is denoise in that region (not CFG). Use this for face / hand / chest fixes without opening Comfy.
+- **generate** — text only. No `image_a`, `image_b`, or mask. Size is width × height (768 / 1024 / 1280), not megapixels from a still. Use this to make a new Klein still from a prompt.
 
 Compose tasks:
 
@@ -27,12 +29,13 @@ Compose tasks:
 - **face_lock** — body/pose/scene from A; face from B
 - **scene** — A is the edit image; B is style/background
 
-Refine does not prepend Edit/Compose role headers.
+Refine and Generate do not prepend Edit/Compose role headers. The user prompt is the whole prompt.
 
 Dummy tests:
 
 - Compose + two unrelated photos + “keep everything the same” must change the output. If it matches old v1 gens, still B is not connected.
 - Refine + last good gen + mask on the left breast + “red X painted on the left breast” at strength 0.4. Pass = X on the breast, face/pose/background stay. Fail = whole image regenerates, or nothing changes.
+- Generate + “a red cube on a white table, studio light” and no stills. Pass = a new image of that. Fail = missing-image error, or the output equals the last Edit/Compose still.
 
 ## Nodes the mapper patches
 
@@ -72,6 +75,18 @@ Refine only (`klein_v2_refine.json`):
 | `17` | `BasicScheduler.denoise` — this is Strength, not CFG |
 | `19` | Sampler starts from the masked canvas latent, not an empty Flux2 latent |
 
+Generate only (`klein_v2_generate.json`):
+
+| Node | Role |
+|------|------|
+| `14` | `EmptyFlux2LatentImage` at request width × height |
+| `17` | `Flux2Scheduler` steps + the same size |
+| `7` | SNOFS LoRA |
+| `8` | Consistency LoRA — removed at runtime when both strengths are 0 |
+| `9` / `10` | Prompt / negative from the API (no leftover widget text) |
+
+There is no `LoadImage`. If the executed Generate graph has a required LoadImage, the job fails. v1 node 145 (muted T2I with a baked prompt) is not used.
+
 If `image_b` is sent on Edit/Compose and the executed graph has fewer than two `LoadImage` nodes, the job fails. If Refine runs without a mask input or without denoise on the scheduler, the job fails.
 
 ## Default sliders
@@ -85,7 +100,8 @@ If `image_b` is sent on Edit/Compose and the executed graph has fewer than two `
 | Steps | 24 (turbo 8) |
 | CFG | 4 (turbo 1) |
 | Scale to MP | 1.0 lanczos |
-| Strength (denoise) | 0.35 on Refine only. Hidden on Edit/Compose. Range 0.15–0.75 |
+| Strength (denoise) | 0.35 on Refine only. Hidden on Edit/Compose/Generate. Range 0.15–0.75 |
+| Size | Generate only. 1:1 1024×1024, 3:4 768×1024, 4:3 1024×768, 16:9 1280×768, 9:16 768×1280 |
 
 ### When to use each Refine strength
 
@@ -103,5 +119,6 @@ Do not raise Strength to rewrite the whole frame. If the face or background move
 - **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.
 - **refine** — prompt only the masked change. Face vs hand/chest chips set Strength and LoRAs together.
+- **generate** — SFW scene = SNOFS 0 / 0. SNOF T2I = SNOFS 0.65 / 0.35. Consistency stays 0. Turbo is 8 steps / CFG 1.
 
-Presets on this tab store sliders + mode + task (+ Strength when Refine). They do not load v1 LoRA stacks or a cached latent.
+Presets on this tab store sliders + mode + task (+ Strength when Refine, + size when Generate). They do not load v1 LoRA stacks or a cached latent.

pages/index.vue

@@ -65,7 +65,9 @@
           </div>
           <h2 class="font-display text-2xl font-bold">Input</h2>
           <p class="text-sm text-zinc-400">{{ studioMode === 'editv2'
-            ? (v2Mode === 'refine'
+            ? (v2Mode === 'generate'
+              ? 'Generate: no reference image. Prompt only. Klein 9B Base text-to-image.'
+              : v2Mode === 'refine'
               ? 'Refine: paint a mask on Still A. Prompt only what should change in the painted area. Strength is denoise in that region.'
               : v2Mode === 'compose'
               ? 'Compose: still A is the person/body. Still B is the face or outfit. Two stills required — no silent one-image fallback.'
@@ -76,6 +78,7 @@
         </div>
 
         <div
+          v-if="!(studioMode === 'editv2' && v2Mode === 'generate')"
           class="relative cursor-pointer rounded-2xl border border-dashed border-white/15 bg-zinc-950/50 p-4 transition hover:border-amber-300/50"
           :class="{ 'border-amber-300/70 bg-amber-400/5': dragging }"
           @dragover.prevent="dragging = true"
@@ -133,6 +136,14 @@
           >
             Refine
           </button>
+          <button
+            type="button"
+            class="rounded-full border px-3 py-1 text-xs font-medium"
+            :class="v2Mode === 'generate' ? 'border-amber-300/70 bg-amber-400/10 text-amber-100' : 'border-white/10 text-zinc-400 hover:text-white'"
+            @click="setV2Mode('generate')"
+          >
+            Generate
+          </button>
           <label v-if="v2Mode === 'compose'" class="ml-auto text-xs">
             <span class="mr-2 text-zinc-500">Task</span>
             <select
@@ -192,7 +203,7 @@
 
         <div>
           <div class="mb-2 flex flex-wrap items-center justify-between gap-2">
-            <label class="text-sm font-medium text-zinc-300">{{ studioMode === 'editv2' && v2Mode === 'refine' ? 'Refine prompt' : studioMode === 'edit' || studioMode === 'editv2' ? 'Edit prompt' : 'Motion & scene prompt' }}</label>
+            <label class="text-sm font-medium text-zinc-300">{{ studioMode === 'editv2' && v2Mode === 'generate' ? 'Generate prompt' : studioMode === 'editv2' && v2Mode === 'refine' ? 'Refine prompt' : studioMode === 'edit' || studioMode === 'editv2' ? 'Edit prompt' : 'Motion & scene prompt' }}</label>
             <div class="flex flex-wrap items-center gap-3">
               <button
                 v-if="identityPrompting"
@@ -276,6 +287,7 @@
               />
             </label>
           </div>
+          <p v-if="studioMode === 'editv2' && v2Mode === 'generate'" class="mt-3 text-xs text-zinc-500">No reference image. Prompt only.</p>
           <div v-if="studioMode === 'editv2' && v2Mode === 'refine'" class="mt-3 space-y-2">
             <p class="text-xs text-zinc-500">Prompt only what should change in the painted area.</p>
             <div class="flex flex-wrap gap-2">
@@ -698,7 +710,40 @@
           />
         </div>
         <div v-else-if="studioMode === 'editv2'" class="space-y-3">
-          <p class="text-xs text-zinc-500">Klein v2 · 9B Base · euler. Edit, Compose, and Refine are separate graphs. Shares the desktop GPU with video.</p>
+          <p class="text-xs text-zinc-500">Klein v2 · 9B Base · euler. Edit, Compose, Refine, and Generate are separate graphs. Shares the desktop GPU with video.</p>
+          <div v-if="v2Mode === 'generate'" class="space-y-2 rounded-2xl border border-white/10 bg-zinc-950/40 p-3">
+            <p class="text-xs text-zinc-400">Size · near 1 MP</p>
+            <div class="flex flex-wrap gap-2">
+              <button
+                v-for="(pair, aspect) in IMAGE_V2_GENERATE_ASPECTS"
+                :key="aspect"
+                type="button"
+                class="rounded-full border px-2.5 py-1 text-xs"
+                :class="v2Width === pair[0] && v2Height === pair[1] ? 'border-amber-300/70 text-amber-100' : 'border-white/10 text-zinc-400'"
+                @click="setGenerateSize(aspect)"
+              >
+                {{ aspect }} · {{ pair[0] }}×{{ pair[1] }}
+              </button>
+            </div>
+            <div class="flex flex-wrap gap-2">
+              <button
+                type="button"
+                class="rounded-full border px-2.5 py-1 text-xs"
+                :class="v2SnofsModel === 0 && v2SnofsClip === 0 ? 'border-amber-300/70 text-amber-100' : 'border-white/10 text-zinc-400'"
+                @click="applyGenerateLook('sfw')"
+              >
+                SFW scene
+              </button>
+              <button
+                type="button"
+                class="rounded-full border px-2.5 py-1 text-xs"
+                :class="v2SnofsModel === IMAGE_V2_GENERATE_SNOFS.snofsModel && v2SnofsClip === IMAGE_V2_GENERATE_SNOFS.snofsClip ? 'border-amber-300/70 text-amber-100' : 'border-white/10 text-zinc-400'"
+                @click="applyGenerateLook('snofs')"
+              >
+                SNOF T2I
+              </button>
+            </div>
+          </div>
           <div v-if="v2Mode === 'refine'" class="space-y-2 rounded-2xl border border-white/10 bg-zinc-950/40 p-3">
             <label class="block text-sm">
               <span class="mb-1 block text-zinc-300">Strength (denoise)</span>
@@ -750,11 +795,11 @@
               <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">
+            <label v-if="v2Mode !== 'generate'" 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">
+            <label v-if="v2Mode !== 'generate'" 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>
@@ -770,7 +815,7 @@
               <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">
+            <label v-if="v2Mode !== 'generate'" 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>
@@ -1327,6 +1372,13 @@
             >
               Use as input still
             </button>
+            <button
+              type="button"
+              class="inline-flex rounded-2xl border border-amber-300/40 px-4 py-2 text-sm text-amber-100 hover:border-amber-300/70"
+              @click="sendToEdit"
+            >
+              Send to Edit
+            </button>
             <button
               type="button"
               class="inline-flex rounded-2xl border border-amber-300/40 px-4 py-2 text-sm text-amber-100 hover:border-amber-300/70"
@@ -1864,6 +1916,11 @@ import {
   IMAGE_V2_DENOISE_MAX,
   IMAGE_V2_DENOISE_MIN,
   IMAGE_V2_DENOISE_STEP,
+  IMAGE_V2_GENERATE_ASPECTS,
+  IMAGE_V2_GENERATE_HEIGHT,
+  IMAGE_V2_GENERATE_SNOFS,
+  IMAGE_V2_GENERATE_SFW,
+  IMAGE_V2_GENERATE_WIDTH,
   IMAGE_V2_REFINE_FACE,
   IMAGE_V2_REFINE_FACE_PROMPT,
   IMAGE_V2_REFINE_HAND,
@@ -1872,7 +1929,9 @@ import {
   IMAGE_V2_SNOFS_MODEL,
   IMAGE_V2_STEPS_DEFAULT,
   clampImageV2Denoise,
+  clampImageV2Size,
   clampImageV2Strength,
+  type ImageV2GenerateAspect,
   type ImageV2Mode,
   type ImageV2PresetSettings,
   type ImageV2Task
@@ -2153,6 +2212,8 @@ const v2Steps = ref(IMAGE_V2_STEPS_DEFAULT)
 const v2Cfg = ref(IMAGE_V2_CFG_DEFAULT)
 const v2Megapixels = ref(1)
 const v2Turbo = ref(false)
+const v2Width = ref(IMAGE_V2_GENERATE_WIDTH)
+const v2Height = ref(IMAGE_V2_GENERATE_HEIGHT)
 const v2Denoise = ref(IMAGE_V2_DENOISE_DEFAULT)
 const refineMaskDirty = ref(false)
 const refinePainter = ref<{ exportPng: () => Promise<Blob | null>; clear: () => void } | null>(null)
@@ -2532,14 +2593,15 @@ const stillBHint = computed(() => {
   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 (!prompt.value.trim() || !folderId.value) return true
+  if (v2Mode.value !== 'generate' && !file.value) return true
   if (v2Mode.value === 'compose' && !editRefFile.value) return true
   if (v2Mode.value === 'refine' && !refineMaskDirty.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 !== 'generate' && !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 (v2Mode.value === 'refine' && !refineMaskDirty.value) return 'Paint a mask on Still A first. Refine will not fall back to Edit.'
   if (!prompt.value.trim()) return 'Write a prompt first.'
@@ -2550,7 +2612,7 @@ const editV2BlockReason = computed(() => {
 })
 const editV2SubmitLabel = computed(() => {
   const occupied = videoBusy.value || editBusy.value || studioJobs.value.some(job => job.status === 'running')
-  const label = v2Mode.value === 'refine' ? 'Refine image' : v2Mode.value === 'compose' ? 'Compose image' : 'Edit image'
+  const label = v2Mode.value === 'generate' ? 'Generate image' : v2Mode.value === 'refine' ? 'Refine image' : v2Mode.value === 'compose' ? 'Compose image' : 'Edit image'
   return occupied ? `Queue ${label.toLowerCase()}` : label
 })
 const generateLabel = computed(() => {
@@ -2608,6 +2670,7 @@ const composedIdentityPrompt = computed(() => {
 })
 const promptPlaceholder = computed(() => {
   if (studioMode.value === 'editv2') {
+    if (v2Mode.value === 'generate') return 'A red cube on a white table, studio light.'
     if (v2Mode.value === 'refine') return 'Prompt only what should change in the painted area.'
     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.'
@@ -3347,17 +3410,19 @@ function currentPresetSnapshot() {
       loraStack: [],
       settings: {
         mode: v2Mode.value,
-        task: v2Mode.value === 'compose' ? v2Task.value : v2Mode.value === 'refine' ? 'refine' : 'scene',
+        task: v2Mode.value === 'compose' ? v2Task.value : v2Mode.value === 'refine' ? 'refine' : v2Mode.value === 'generate' ? 't2i' : '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),
+        consistencyModel: v2Mode.value === 'generate' ? 0 : clampImageV2Strength(v2ConsistencyModel.value, IMAGE_V2_CONSISTENCY_MODEL),
+        consistencyClip: v2Mode.value === 'generate' ? 0 : 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),
+        megapixels: v2Mode.value === 'generate' ? undefined : clampImageScaleMegapixels(v2Megapixels.value, 1),
         turbo: v2Turbo.value === true,
-        strength: v2Mode.value === 'refine' ? clampImageV2Denoise(v2Denoise.value) : undefined
+        strength: v2Mode.value === 'refine' ? clampImageV2Denoise(v2Denoise.value) : undefined,
+        width: v2Mode.value === 'generate' ? v2Width.value : undefined,
+        height: v2Mode.value === 'generate' ? v2Height.value : undefined
       }
     }
   }
@@ -3396,18 +3461,23 @@ function applyGenerationPreset(preset: GenerationPreset) {
   const skipped = preset.loraStack.length - stack.length
   if (preset.kind === 'imagev2') {
     const settings = preset.settings as ImageV2PresetSettings
-    v2Mode.value = settings.mode === 'compose' ? 'compose' : settings.mode === 'refine' ? 'refine' : 'edit'
-    v2Task.value = v2Mode.value === 'compose' ? (settings.task || 'scene') : v2Mode.value === 'refine' ? 'refine' : 'scene'
+    v2Mode.value = settings.mode === 'compose' ? 'compose' : settings.mode === 'refine' ? 'refine' : settings.mode === 'generate' ? 'generate' : 'edit'
+    v2Task.value = v2Mode.value === 'compose' ? (settings.task || 'scene') : v2Mode.value === 'refine' ? 'refine' : v2Mode.value === 'generate' ? 't2i' : '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)
+    v2ConsistencyModel.value = v2Mode.value === 'generate' ? 0 : clampImageV2Strength(settings.consistencyModel, IMAGE_V2_CONSISTENCY_MODEL)
+    v2ConsistencyClip.value = v2Mode.value === 'generate' ? 0 : 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
     if (v2Mode.value === 'refine') v2Denoise.value = clampImageV2Denoise(settings.strength)
+    if (v2Mode.value === 'generate') {
+      const size = clampImageV2Size(settings.width, settings.height)
+      v2Width.value = size.width
+      v2Height.value = size.height
+    }
     void ensureRefineCanvas()
     imageV2PresetId.value = preset.id
     imageV2PresetName.value = preset.name
@@ -5196,12 +5266,35 @@ async function editImage() {
 }
 
 function setV2Mode(mode: ImageV2Mode) {
+  const previous = v2Mode.value
   v2Mode.value = mode
   if (mode === 'edit') v2Task.value = 'scene'
   if (mode === 'refine') {
     v2Task.value = 'refine'
     void ensureRefineCanvas()
   }
+  if (mode === 'generate') {
+    v2Task.value = 't2i'
+    v2ConsistencyModel.value = 0
+    v2ConsistencyClip.value = 0
+  } else if (previous === 'generate' && v2ConsistencyModel.value === 0 && v2ConsistencyClip.value === 0) {
+    v2ConsistencyModel.value = IMAGE_V2_CONSISTENCY_MODEL
+    v2ConsistencyClip.value = IMAGE_V2_CONSISTENCY_CLIP
+  }
+}
+
+function setGenerateSize(aspect: ImageV2GenerateAspect) {
+  const [width, height] = IMAGE_V2_GENERATE_ASPECTS[aspect]
+  v2Width.value = width
+  v2Height.value = height
+}
+
+function applyGenerateLook(kind: 'sfw' | 'snofs') {
+  const look = kind === 'sfw' ? IMAGE_V2_GENERATE_SFW : IMAGE_V2_GENERATE_SNOFS
+  v2SnofsModel.value = look.snofsModel
+  v2SnofsClip.value = look.snofsClip
+  v2ConsistencyModel.value = 0
+  v2ConsistencyClip.value = 0
 }
 
 function applyRefinePreset(kind: 'face' | 'hand' | 'heavy') {
@@ -5238,6 +5331,19 @@ async function ensureRefineCanvas() {
   }
 }
 
+async function sendToEdit() {
+  if (!editResultUrl.value) return
+  studioMode.value = 'editv2'
+  outputFocus.value = 'edit'
+  try {
+    await useEditAsInput()
+    setV2Mode('edit')
+    toast('Output is Still A. Write an edit prompt.')
+  } catch {
+    toast('Could not send that still to Edit')
+  }
+}
+
 async function sendToRefine() {
   if (!editResultUrl.value) return
   studioMode.value = 'editv2'
@@ -5268,8 +5374,8 @@ async function editImageV2() {
   try {
     const body = new FormData()
     body.append('mode', v2Mode.value)
-    body.append('task', v2Mode.value === 'refine' ? 'refine' : v2Mode.value === 'compose' ? v2Task.value : 'scene')
-    body.append('image_a', file.value as File)
+    body.append('task', v2Mode.value === 'generate' ? 't2i' : v2Mode.value === 'refine' ? 'refine' : v2Mode.value === 'compose' ? v2Task.value : 'scene')
+    if (v2Mode.value !== 'generate' && file.value) body.append('image_a', file.value)
     if (v2Mode.value === 'compose' && editRefFile.value) body.append('image_b', editRefFile.value)
     if (v2Mode.value === 'refine') {
       const maskBlob = await refinePainter.value?.exportPng()
@@ -5280,15 +5386,20 @@ async function editImageV2() {
       body.append('mask', new File([maskBlob], 'refine-mask.png', { type: 'image/png' }))
       body.append('strength', String(clampImageV2Denoise(v2Denoise.value)))
     }
+    if (v2Mode.value === 'generate') {
+      const size = clampImageV2Size(v2Width.value, v2Height.value)
+      body.append('width', String(size.width))
+      body.append('height', String(size.height))
+    }
     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('consistency_model', String(v2Mode.value === 'generate' ? 0 : clampImageV2Strength(v2ConsistencyModel.value, IMAGE_V2_CONSISTENCY_MODEL)))
+    body.append('consistency_clip', String(v2Mode.value === 'generate' ? 0 : 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)))
+    if (v2Mode.value !== 'generate') 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)
@@ -5328,7 +5439,7 @@ async function editImageV2() {
     editChainLabel.value = ''
     editOverallProgress.value = 0
     editCompletedChainStep.value = 0
-    editActiveChainPlan.value = [{ label: v2Mode.value === 'refine' ? 'Refine' : v2Mode.value === 'compose' ? 'Compose' : 'Edit', prompt: prompt.value.trim() }]
+    editActiveChainPlan.value = [{ label: v2Mode.value === 'generate' ? 'Generate' : v2Mode.value === 'refine' ? 'Refine' : v2Mode.value === 'compose' ? 'Compose' : 'Edit', prompt: prompt.value.trim() }]
     startTimer('edit')
     editDownloadName.value = downloadName
     editJobId.value = started.jobId

server/api/v2/generate.post.ts

@@ -13,6 +13,7 @@ import {
   IMAGE_V2_TURBO_STEPS,
   IMAGE_V2_DENOISE_DEFAULT,
   clampImageV2Denoise,
+  clampImageV2Size,
   clampImageV2Strength,
   parseImageV2Mode,
   parseImageV2Task,
@@ -109,9 +110,12 @@ export default defineEventHandler(async (event) => {
 
   const mode = parseImageV2Mode(fields.mode)
   if (!mode) {
-    throw createError({ statusCode: 400, statusMessage: 'mode must be edit, compose, or refine' })
+    throw createError({ statusCode: 400, statusMessage: 'mode must be edit, compose, refine, or generate' })
   }
-  const task = parseImageV2Task(fields.task, mode === 'refine' ? 'refine' : 'scene')
+  const task = parseImageV2Task(
+    fields.task,
+    mode === 'generate' ? 't2i' : mode === 'refine' ? 'refine' : 'scene'
+  )
   const prompt = String(fields.prompt || '').trim()
   if (!prompt) {
     throw createError({ statusCode: 400, statusMessage: 'A prompt is required' })
@@ -131,11 +135,11 @@ export default defineEventHandler(async (event) => {
   }
 
   const ownerKey = libraryOwnerKey(event)
-  const imageA = await resolveImageRef(ownerKey, fields.image_a, uploadedA)
-  const imageB = mode === 'refine' ? null : await resolveImageRef(ownerKey, fields.image_b, uploadedB)
+  const imageA = mode === 'generate' ? null : await resolveImageRef(ownerKey, fields.image_a, uploadedA)
+  const imageB = mode === 'refine' || mode === 'generate' ? null : await resolveImageRef(ownerKey, fields.image_b, uploadedB)
   const mask = mode === 'refine' ? await resolveImageRef(ownerKey, fields.mask, uploadedMask) : null
 
-  if (!imageA) {
+  if (mode !== 'generate' && !imageA) {
     throw createError({ statusCode: 400, statusMessage: 'image_a is required' })
   }
   if (mode === 'refine' && !mask) {
@@ -173,17 +177,25 @@ export default defineEventHandler(async (event) => {
   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 megapixels = mode === 'generate' ? 0 : 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 generateSize = clampImageV2Size(fields.width, fields.height, fields.aspect)
+  const size = mode === 'generate'
+    ? { width: generateSize.width, height: generateSize.height }
+    : imageDimensions(imageA!.data)
   const clipName = String(fields.name || '').trim().slice(0, 80)
   const v2Mode = mode as ImageV2Mode
-  const v2Task = (mode === 'refine' ? 'refine' : mode === 'edit' ? 'scene' : task) as ImageV2Task
+  const v2Task = (
+    mode === 'generate' ? 't2i' : mode === 'refine' ? 'refine' : mode === 'edit' ? 'scene' : task
+  ) as ImageV2Task
   const refineStrength = mode === 'refine' ? clampImageV2Denoise(fields.strength, IMAGE_V2_DENOISE_DEFAULT) : undefined
+  const generateConsistencyModel = clampImageV2Strength(fields.consistency_model, mode === 'generate' ? 0 : IMAGE_V2_CONSISTENCY_MODEL)
+  const generateConsistencyClip = clampImageV2Strength(fields.consistency_clip, mode === 'generate' ? 0 : IMAGE_V2_CONSISTENCY_CLIP)
 
-  const still = await rememberInputStill({
+  const still = imageA
+    ? await rememberInputStill({
       ownerKey,
       folderId,
       filename: imageA.filename,
@@ -192,6 +204,7 @@ export default defineEventHandler(async (event) => {
       height: size?.height || 0,
       hideInput
     })
+    : null
   const savedRef = imageB
     ? await rememberInputStill({
       ownerKey,
@@ -219,7 +232,7 @@ export default defineEventHandler(async (event) => {
       prompt,
       name: clipName,
       folderId,
-      aspect: 'auto',
+      aspect: mode === 'generate' ? (generateSize.aspect || '1:1') : 'auto',
       width: size?.width || 0,
       height: size?.height || 0,
       steps,
@@ -244,15 +257,15 @@ export default defineEventHandler(async (event) => {
       negative: String(fields.negative || '').trim(),
       referenceStillId: savedRef?.id,
       referenceStillFilename: savedRef?.filename,
-      scaleToTotalPixels: true,
+      scaleToTotalPixels: mode !== 'generate',
       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),
+      consistencyModel: generateConsistencyModel,
+      consistencyClip: generateConsistencyClip,
       maskStillId: savedMask?.id,
       maskStillFilename: savedMask?.filename,
       refineStrength
@@ -275,7 +288,11 @@ export default defineEventHandler(async (event) => {
     cfg,
     mode: v2Mode,
     task: v2Task,
-    workflow: v2Mode === 'refine' ? 'klein_v2_refine.json' : v2Mode === 'compose' ? 'klein_v2_compose.json' : 'klein_v2_edit.json',
+    workflow: v2Mode === 'generate'
+      ? 'klein_v2_generate.json'
+      : v2Mode === 'refine' ? 'klein_v2_refine.json' : v2Mode === 'compose' ? 'klein_v2_compose.json' : 'klein_v2_edit.json',
+    width: size?.width || 0,
+    height: size?.height || 0,
     strength: refineStrength,
     hideThumbnail,
     folderLocked

server/assets/klein_v2_generate.json

@@ -0,0 +1,127 @@
+{
+  "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,
+      "strength_clip": 0,
+      "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" }
+  },
+  "14": {
+    "inputs": {
+      "width": 1024,
+      "height": 1024,
+      "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": 1024,
+      "height": 1024
+    },
+    "class_type": "Flux2Scheduler",
+    "_meta": { "title": "Flux2Scheduler" }
+  },
+  "18": {
+    "inputs": {
+      "cfg": 4,
+      "model": ["8", 0],
+      "positive": ["9", 0],
+      "negative": ["10", 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-generate",
+      "images": ["20", 0]
+    },
+    "class_type": "SaveImage",
+    "_meta": { "title": "Save Image" }
+  }
+}

server/utils/generationPresets.ts

@@ -23,6 +23,7 @@ import {
   IMAGE_V2_STEPS_DEFAULT,
   IMAGE_V2_DENOISE_DEFAULT,
   clampImageV2Denoise,
+  clampImageV2Size,
   clampImageV2Strength,
   parseImageV2Mode,
   parseImageV2Task,
@@ -116,17 +117,19 @@ function sanitizeImageV2Settings(raw: unknown): ImageV2PresetSettings {
   const turbo = rec.turbo === true
   return {
     mode,
-    task: mode === 'compose' ? parseImageV2Task(rec.task, 'scene') : mode === 'refine' ? 'refine' : 'scene',
+    task: mode === 'compose' ? parseImageV2Task(rec.task, 'scene') : mode === 'refine' ? 'refine' : mode === 'generate' ? 't2i' : '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),
+    consistencyModel: clampImageV2Strength(rec.consistencyModel ?? rec.consistency_model, mode === 'generate' ? 0 : IMAGE_V2_CONSISTENCY_MODEL),
+    consistencyClip: clampImageV2Strength(rec.consistencyClip ?? rec.consistency_clip, mode === 'generate' ? 0 : 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),
+    megapixels: mode === 'generate' ? undefined : clampImageScaleMegapixels(rec.megapixels ?? rec.scaleMegapixels, 1),
     turbo,
-    strength: mode === 'refine' ? clampImageV2Denoise(rec.strength, IMAGE_V2_DENOISE_DEFAULT) : undefined
+    strength: mode === 'refine' ? clampImageV2Denoise(rec.strength, IMAGE_V2_DENOISE_DEFAULT) : undefined,
+    width: mode === 'generate' ? clampImageV2Size(rec.width, rec.height, rec.aspect).width : undefined,
+    height: mode === 'generate' ? clampImageV2Size(rec.width, rec.height, rec.aspect).height : undefined
   }
 }
 

server/utils/imageChainV2.ts

@@ -13,7 +13,7 @@ import type { EditImageFile } from '~/server/utils/imageChain'
 export type EditV2RunParams = {
   mode: ImageV2Mode
   task: ImageV2Task
-  image: EditImageFile
+  image?: EditImageFile | null
   reference: EditImageFile | null
   mask?: EditImageFile | null
   prompt: string
@@ -27,6 +27,9 @@ export type EditV2RunParams = {
   consistencyClip: number
   megapixels: number
   strength?: number
+  width?: number
+  height?: number
+  turbo?: boolean
 }
 
 export async function runEditV2(job: Job, params: EditV2RunParams) {
@@ -41,6 +44,9 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
   if (params.mode === 'edit' && params.reference) {
     throw new Error('Edit mode takes one image. Use Compose for two stills.')
   }
+  if (params.mode !== 'generate' && !params.image) {
+    throw new Error('This v2 mode requires still A.')
+  }
 
   try {
     await ensureComfyReady((status) => {
@@ -55,15 +61,18 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
     job.imageComfyHost = getComfyHost()
 
     await withImageComfyHost(job.imageComfyHost, async () => {
+      const generate = params.mode === 'generate'
       emitChainJob(job, {
         type: 'status',
-        message: params.mode === 'refine'
+        message: generate
+          ? 'Queueing Klein v2 generate on Beast...'
+          : params.mode === 'refine'
             ? 'Uploading canvas and mask to Beast...'
             : params.mode === 'compose' ? 'Uploading stills A and B to Beast...' : 'Uploading still A to Beast...',
-        progress: 8
+        progress: generate ? 12 : 8
       })
-      const uploaded = await uploadImage(params.image, job.id)
-      const uploadedRef = params.mode === 'refine'
+      const uploaded = generate || !params.image ? null : await uploadImage(params.image, job.id)
+      const uploadedRef = generate || params.mode === 'refine'
         ? null
         : params.reference
           ? await uploadImage({
@@ -71,7 +80,7 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
             filename: `ref_${params.reference.filename || 'image_b.png'}`
           }, job.id)
           : null
-      const uploadedMask = params.mode === 'refine' && params.mask
+      const uploadedMask = !generate && params.mode === 'refine' && params.mask
         ? await uploadImage({
           ...params.mask,
           filename: `mask_${params.mask.filename || 'refine-mask.png'}`
@@ -79,6 +88,7 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
         : null
       if (job.status === 'cancelled') throw new Error('Job interrupted.')
 
+      if (!generate) {
         emitChainJob(job, {
           type: 'status',
           message: params.mode === 'refine'
@@ -88,14 +98,15 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
               : 'Queueing Klein v2 edit on Beast...',
           progress: 12
         })
+      }
       await ensureComfyLoraNames('image')
-      await assertImageScaleToTotalPixelsNode()
+      if (!generate) await assertImageScaleToTotalPixelsNode()
       const built = buildImageV2Workflow({
         mode: params.mode,
         task: params.task,
         prompt: params.prompt,
         negative: params.negative,
-        imageAName: uploaded.name,
+        imageAName: uploaded?.name,
         imageBName: uploadedRef?.name,
         maskName: uploadedMask?.name,
         strength: params.strength,
@@ -107,6 +118,9 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
         cfg: params.cfg,
         seed: params.seed,
         megapixels: params.megapixels,
+        width: params.width,
+        height: params.height,
+        turbo: params.turbo === true,
         filenamePrefix: `aigen_v2_${job.id.slice(0, 8)}`
       })
       const queued = await queuePrompt(built.graph, job.clientId)
@@ -154,8 +168,8 @@ export async function runEditV2(job: Job, params: EditV2RunParams) {
       job.stillId = still?.id
       await purgeComfyArtifacts({
         video: { filename: output.filename, subfolder: output.subfolder, type: output.type },
-        imageName: uploaded.name,
-        imageSubfolder: uploaded.subfolder,
+        imageName: uploaded?.name,
+        imageSubfolder: uploaded?.subfolder,
         extraImageNames: [uploadedRef?.name, uploadedMask?.name].filter((name): name is string => Boolean(name)),
         promptId: job.promptId
       })

server/utils/imageWorkflowV2.ts

@@ -1,6 +1,7 @@
 import editTemplate from '../assets/klein_v2_edit.json'
 import composeTemplate from '../assets/klein_v2_compose.json'
 import refineTemplate from '../assets/klein_v2_refine.json'
+import generateTemplate from '../assets/klein_v2_generate.json'
 import { IMAGE_SCALE_TO_TOTAL_PIXELS } from '~/server/utils/comfy'
 import { cachedComfyLoraNames } from '~/server/utils/loras'
 import { resolveComfyLoraName, loraIdentityKey } from '~/utils/loras'
@@ -13,7 +14,10 @@ import {
   IMAGE_V2_SNOFS_LORA,
   IMAGE_V2_SNOFS_MODEL,
   IMAGE_V2_DENOISE_DEFAULT,
+  IMAGE_V2_GENERATE_HEIGHT,
+  IMAGE_V2_GENERATE_WIDTH,
   clampImageV2Denoise,
+  clampImageV2Size,
   clampImageV2Strength,
   composeImageV2Prompt,
   type ImageV2Mode,
@@ -41,16 +45,20 @@ const CONSISTENCY = '8'
 export const IMAGE_V2_EDIT_WORKFLOW = 'klein_v2_edit.json'
 export const IMAGE_V2_COMPOSE_WORKFLOW = 'klein_v2_compose.json'
 export const IMAGE_V2_REFINE_WORKFLOW = 'klein_v2_refine.json'
+export const IMAGE_V2_GENERATE_WORKFLOW = 'klein_v2_generate.json'
 
 export interface ImageV2BuildParams {
   mode: ImageV2Mode
   task: ImageV2Task
   prompt: string
   negative?: string
-  imageAName: string
+  imageAName?: string
   imageBName?: string
   maskName?: string
   strength?: number
+  width?: number
+  height?: number
+  turbo?: boolean
   snofsModel?: number
   snofsClip?: number
   consistencyModel?: number
@@ -102,8 +110,30 @@ function graphHasMaskInput(graph: WorkflowGraph) {
   })
 }
 
+function bypassLoraNode(graph: WorkflowGraph, id: string, modelFrom: string, clipFrom: string) {
+  delete graph[id]
+  for (const node of Object.values(graph)) {
+    for (const [key, value] of Object.entries(node.inputs)) {
+      if (!Array.isArray(value) || value[0] !== id) continue
+      node.inputs[key] = value[1] === 1 ? [clipFrom, 1] : [modelFrom, 0]
+    }
+  }
+}
+
 export function assertImageV2Graph(graph: WorkflowGraph, mode: ImageV2Mode, imageBName?: string) {
   const loaders = loadImageNames(graph)
+  if (mode === 'generate') {
+    if (loaders.length) {
+      throw createError({
+        statusCode: 500,
+        statusMessage: 'Generate graph has a required LoadImage. Refusing to run an edit fallback.'
+      })
+    }
+    const latent = Object.values(graph).find(node => node.class_type === 'EmptyFlux2LatentImage')
+    if (!latent) {
+      throw createError({ statusCode: 500, statusMessage: 'Generate graph is missing EmptyFlux2LatentImage.' })
+    }
+  }
   if (mode === 'refine') {
     const mask = graph[LOAD_MASK]
     if (!mask || mask.class_type !== 'LoadImage' || !String(mask.inputs.image || '').trim()) {
@@ -140,7 +170,7 @@ export function assertImageV2Graph(graph: WorkflowGraph, mode: ImageV2Mode, imag
       })
     }
   }
-  if (mode !== 'refine' && imageBName && loaders.length < 2) {
+  if (mode !== 'refine' && mode !== 'generate' && imageBName && loaders.length < 2) {
     throw createError({
       statusCode: 500,
       statusMessage: 'image_b was sent but the executed graph has no second image input.'
@@ -161,43 +191,65 @@ export function assertImageV2Graph(graph: WorkflowGraph, mode: ImageV2Mode, imag
 export function buildImageV2Workflow(params: ImageV2BuildParams) {
   const compose = params.mode === 'compose'
   const refine = params.mode === 'refine'
+  const generate = params.mode === 'generate'
   if (refine && !String(params.maskName || '').trim()) {
     throw createError({ statusCode: 400, statusMessage: 'Refine requires a mask. Refusing to fall back to Edit.' })
   }
-  const graph = structuredClone(refine ? refineTemplate : compose ? composeTemplate : editTemplate) as WorkflowGraph
+  const graph = structuredClone(
+    generate ? generateTemplate : refine ? refineTemplate : 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 consistencyModel = clampImageV2Strength(
+    params.consistencyModel,
+    generate ? 0 : IMAGE_V2_CONSISTENCY_MODEL
+  )
+  const consistencyClip = clampImageV2Strength(
+    params.consistencyClip,
+    generate ? 0 : IMAGE_V2_CONSISTENCY_CLIP
+  )
   const steps = clampImageSteps(params.steps, 24)
   const cfg = clampImageCfg(params.cfg, 4)
   const strength = refine ? clampImageV2Denoise(params.strength, IMAGE_V2_DENOISE_DEFAULT) : undefined
-  const workflowFile = refine ? IMAGE_V2_REFINE_WORKFLOW : compose ? IMAGE_V2_COMPOSE_WORKFLOW : IMAGE_V2_EDIT_WORKFLOW
+  const size = generate ? clampImageV2Size(params.width, params.height) : null
+  const workflowFile = generate
+    ? IMAGE_V2_GENERATE_WORKFLOW
+    : refine ? IMAGE_V2_REFINE_WORKFLOW : compose ? IMAGE_V2_COMPOSE_WORKFLOW : IMAGE_V2_EDIT_WORKFLOW
 
-  setInput(graph, LOAD_A, 'image', params.imageAName)
+  if (!generate) setInput(graph, LOAD_A, 'image', params.imageAName || '')
   if (compose) setInput(graph, LOAD_B, 'image', params.imageBName || '')
   if (refine) {
     setInput(graph, LOAD_MASK, 'image', params.maskName || '')
     setInput(graph, SCHEDULER_DENOISE, 'denoise', strength)
   }
+  if (generate && size) {
+    setInput(graph, '14', 'width', size.width)
+    setInput(graph, '14', 'height', size.height)
+    setInput(graph, SCHEDULER, 'width', size.width)
+    setInput(graph, SCHEDULER, 'height', size.height)
+  }
   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, SAVE, 'filename_prefix', params.filenamePrefix || (generate ? 'aigen-v2-generate' : 'aigen-v2'))
+  if (!generate) 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)
+  if (generate && consistencyModel <= 0 && consistencyClip <= 0) {
+    bypassLoraNode(graph, CONSISTENCY, SNOFS, SNOFS)
+  } else {
     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, refine ? undefined : params.imageBName)
+  assertImageV2Graph(graph, params.mode, generate || refine ? undefined : params.imageBName)
 
   const loaders = loadImageNames(graph)
   console.log(JSON.stringify({
@@ -205,18 +257,22 @@ export function buildImageV2Workflow(params: ImageV2BuildParams) {
     workflow: workflowFile,
     mode: params.mode,
     task: params.task,
-    canvas: { id: LOAD_A, file: graph[LOAD_A]?.inputs.image },
+    canvas: generate ? undefined : { id: LOAD_A, file: graph[LOAD_A]?.inputs.image },
     mask: refine ? { id: LOAD_MASK, file: graph[LOAD_MASK]?.inputs.image } : undefined,
+    size: generate ? { width: size?.width ?? IMAGE_V2_GENERATE_WIDTH, height: size?.height ?? IMAGE_V2_GENERATE_HEIGHT } : undefined,
     strength,
+    turbo: params.turbo === true,
     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 }
+      consistency: graph[CONSISTENCY]
+        ? { name: graph[CONSISTENCY]?.inputs.lora_name, model: consistencyModel, clip: consistencyClip }
+        : { loaded: false, model: 0, clip: 0 }
     },
     steps,
     cfg,
     seed: params.seed,
-    megapixels: clampImageScaleMegapixels(params.megapixels ?? 1)
+    megapixels: generate ? undefined : clampImageScaleMegapixels(params.megapixels ?? 1)
   }))
 
   return { graph, workflowFile, loaders, prompt, strength }
@@ -238,6 +294,7 @@ export const IMAGE_V2_NODE_LABELS: Record<string, string> = {
   '11': 'Encoding image A',
   '24': 'Encoding image B',
   '33': 'Applying mask',
+  '14': 'Building empty Klein latent',
   '19': 'Sampling Klein v2',
   '20': 'Decoding still',
   '21': 'Saving still'

server/utils/studioQueue.ts

@@ -52,8 +52,8 @@ export interface StudioJobPayload {
   scaleToTotalPixels?: boolean
   scaleMegapixels?: number
   imagePipeline?: 'v1' | 'v2'
-  v2Mode?: 'edit' | 'compose' | 'refine'
-  v2Task?: 'scene' | 'identity' | 'outfit' | 'face_lock' | 'refine'
+  v2Mode?: 'edit' | 'compose' | 'refine' | 'generate'
+  v2Task?: 'scene' | 'identity' | 'outfit' | 'face_lock' | 'refine' | 't2i'
   snofsModel?: number
   snofsClip?: number
   consistencyModel?: number
@@ -679,14 +679,17 @@ async function startStudioEditJob(item: StudioJob) {
     const { runEditV2 } = await import('~/server/utils/imageChainV2')
 
     const payload = item.payload
-    if (!payload.stillId || !existsSync(stillPath(item.ownerKey, payload.stillId))) {
+    const generate = payload.imagePipeline === 'v2' && payload.v2Mode === 'generate'
+    if (!generate && (!payload.stillId || !existsSync(stillPath(item.ownerKey, payload.stillId)))) {
       throw new Error('The input still is missing from the library')
     }
-    const image = {
+    const image = payload.stillId && existsSync(stillPath(item.ownerKey, payload.stillId))
+      ? {
         filename: payload.stillFilename || 'still.png',
         data: readFileSync(stillPath(item.ownerKey, payload.stillId)),
         type: 'image/png'
       }
+      : null
     const refId = payload.referenceStillId
     const reference = refId && existsSync(stillPath(item.ownerKey, refId))
       ? {
@@ -697,7 +700,13 @@ async function startStudioEditJob(item: StudioJob) {
       : null
 
     if (payload.imagePipeline === 'v2') {
-      const mode = payload.v2Mode === 'compose' ? 'compose' : payload.v2Mode === 'refine' ? 'refine' : 'edit'
+      const mode = payload.v2Mode === 'compose'
+        ? 'compose'
+        : payload.v2Mode === 'refine'
+          ? 'refine'
+          : payload.v2Mode === 'generate'
+            ? 'generate'
+            : 'edit'
       const maskId = payload.maskStillId
       const mask = mode === 'refine' && maskId && existsSync(stillPath(item.ownerKey, maskId))
         ? {
@@ -745,8 +754,8 @@ async function startStudioEditJob(item: StudioJob) {
       void runEditV2(live, {
         mode,
         task: payload.v2Task || 'scene',
-        image,
-        reference: mode === 'refine' ? null : reference,
+        image: mode === 'generate' ? null : image,
+        reference: mode === 'refine' || mode === 'generate' ? null : reference,
         mask,
         prompt: payload.prompt,
         negative: payload.negative || '',
@@ -755,10 +764,13 @@ async function startStudioEditJob(item: StudioJob) {
         cfg: payload.cfg,
         snofsModel: payload.snofsModel ?? 0.65,
         snofsClip: payload.snofsClip ?? 0.35,
-        consistencyModel: payload.consistencyModel ?? 0.7,
-        consistencyClip: payload.consistencyClip ?? 0.7,
+        consistencyModel: payload.consistencyModel ?? (mode === 'generate' ? 0 : 0.7),
+        consistencyClip: payload.consistencyClip ?? (mode === 'generate' ? 0 : 0.7),
         megapixels: payload.scaleMegapixels ?? 1,
-        strength: payload.refineStrength
+        strength: payload.refineStrength,
+        width: payload.width,
+        height: payload.height,
+        turbo: payload.turbo === true
       }).catch((error) => {
         const message = error instanceof Error ? error.message : String(error)
         if (live && live.status !== 'error' && live.status !== 'cancelled' && live.status !== 'deferred') {

utils/imageV2.ts

@@ -1,5 +1,5 @@
-export const IMAGE_V2_MODES = ['edit', 'compose', 'refine'] as const
-export const IMAGE_V2_TASKS = ['scene', 'identity', 'outfit', 'face_lock', 'refine'] as const
+export const IMAGE_V2_MODES = ['edit', 'compose', 'refine', 'generate'] as const
+export const IMAGE_V2_TASKS = ['scene', 'identity', 'outfit', 'face_lock', 'refine', 't2i'] as const
 
 export type ImageV2Mode = (typeof IMAGE_V2_MODES)[number]
 export type ImageV2Task = (typeof IMAGE_V2_TASKS)[number]
@@ -38,8 +38,21 @@ export const IMAGE_V2_REFINE_HAND = {
 }
 export const IMAGE_V2_REFINE_HEAVY = { strength: 0.55 }
 export const IMAGE_V2_REFINE_FACE_PROMPT = 'same face as the canvas, same glasses, same cheeks and jaw. Change only the face in the masked area.'
+export const IMAGE_V2_SIZE_SIDES = [768, 1024, 1280] as const
+export const IMAGE_V2_GENERATE_WIDTH = 1024
+export const IMAGE_V2_GENERATE_HEIGHT = 1024
+export const IMAGE_V2_GENERATE_ASPECTS = {
+  '1:1': [1024, 1024],
+  '3:4': [768, 1024],
+  '4:3': [1024, 768],
+  '16:9': [1280, 768],
+  '9:16': [768, 1280]
+} as const
+export type ImageV2GenerateAspect = keyof typeof IMAGE_V2_GENERATE_ASPECTS
+export const IMAGE_V2_GENERATE_SFW = { snofsModel: 0, snofsClip: 0 }
+export const IMAGE_V2_GENERATE_SNOFS = { snofsModel: 0.65, snofsClip: 0.35 }
 
-export const IMAGE_V2_ROLE_HEADERS: Record<Exclude<ImageV2Task, 'scene' | 'refine'>, string> = {
+export const IMAGE_V2_ROLE_HEADERS: Record<Exclude<ImageV2Task, 'scene' | 'refine' | 't2i'>, 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.'
@@ -69,10 +82,36 @@ export function clampImageV2Denoise(raw: unknown, fallback = IMAGE_V2_DENOISE_DE
   return Math.min(IMAGE_V2_DENOISE_MAX, Math.max(IMAGE_V2_DENOISE_MIN, Math.round(snapped * 100) / 100))
 }
 
+export function snapImageV2Side(raw: unknown, fallback = IMAGE_V2_GENERATE_WIDTH) {
+  const value = Number(raw)
+  if (!Number.isFinite(value)) return fallback
+  return IMAGE_V2_SIZE_SIDES.reduce((best, side) => (
+    Math.abs(side - value) < Math.abs(best - value) ? side : best
+  ), IMAGE_V2_SIZE_SIDES[0])
+}
+
+export function parseImageV2Aspect(raw: unknown): ImageV2GenerateAspect | null {
+  const value = String(raw || '').trim()
+  return value in IMAGE_V2_GENERATE_ASPECTS ? value as ImageV2GenerateAspect : null
+}
+
+export function clampImageV2Size(width: unknown, height: unknown, aspect?: unknown) {
+  const preset = parseImageV2Aspect(aspect)
+  if (preset) {
+    const [w, h] = IMAGE_V2_GENERATE_ASPECTS[preset]
+    return { width: w, height: h, aspect: preset }
+  }
+  return {
+    width: snapImageV2Side(width, IMAGE_V2_GENERATE_WIDTH),
+    height: snapImageV2Side(height, IMAGE_V2_GENERATE_HEIGHT),
+    aspect: null as ImageV2GenerateAspect | null
+  }
+}
+
 export function composeImageV2Prompt(mode: ImageV2Mode, task: ImageV2Task, prompt: string) {
   const body = String(prompt || '').trim()
-  if (mode === 'refine' || mode !== 'compose' || task === 'scene') return body
-  const header = IMAGE_V2_ROLE_HEADERS[task as Exclude<ImageV2Task, 'scene' | 'refine'>]
+  if (mode === 'generate' || mode === 'refine' || mode !== 'compose' || task === 'scene' || task === 't2i') return body
+  const header = IMAGE_V2_ROLE_HEADERS[task as Exclude<ImageV2Task, 'scene' | 'refine' | 't2i'>]
   return header ? `${header}\n\n${body}` : body
 }
 
@@ -89,4 +128,6 @@ export type ImageV2PresetSettings = {
   megapixels?: number
   turbo?: boolean
   strength?: number
+  width?: number
+  height?: number
 }