Always prepend the Comfy identity prefix and keep refs on every shot.

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

pages/index.vue

@@ -109,7 +109,7 @@
           </div>
           <p v-if="shotScriptMode && studioMode === 'video'" class="mb-2 text-xs text-zinc-500">
             Paste a block. Text before the first marker is shot 1. Put <span class="font-medium text-zinc-300">shot 2</span>, <span class="font-medium text-zinc-300">shot 3</span> on their own lines to queue extensions — a line like <span class="font-medium text-zinc-300">Shot Type</span> will not start a new shot. Every shot uses the duration slider.
-            <span v-if="identityPrompting"> Identity boilerplate is prepended to the first shot only, then concatenated into Reference to Video.</span>
+            <span v-if="identityPrompting"> The Comfy identity prefix is prepended to every shot, and the same Picture 1 / extra stills are sent with each shot.</span>
           </p>
           <p v-else-if="identityPrompting" class="mb-2 text-xs text-zinc-500">
             Write camera, action, and audio. That is the dynamic action prompt. We concatenate the <span class="font-medium text-zinc-300">&lt;Picture 1&gt;</span> identity prefix in front of it and send the result to <span class="font-medium text-zinc-300">Reference to Video</span>, not Image to Video.
@@ -140,7 +140,7 @@
                 {{ index === 0 ? 'Initial' : `Extension ${index}` }}
                 <span class="font-normal text-zinc-500"> · shot {{ shot.n }} · {{ formatClipDuration(duration) }}</span>
               </p>
-              <p class="mt-1 whitespace-pre-wrap text-xs text-zinc-300">{{ shot.prompt }}</p>
+              <p class="mt-1 whitespace-pre-wrap text-xs text-zinc-300">{{ identityPrompting ? buildIdentityPrompt(shot.prompt, extraIdentityPictures) : shot.prompt }}</p>
             </div>
           </div>
         </div>
@@ -321,7 +321,7 @@
                 </button>
               </div>
             </div>
-            <p v-if="useIdentityRefs" class="text-xs text-zinc-500">Picture 1 is the start still. Extra stills map to Picture 2–5. Empty extra slots reuse the start still so Comfy does not error. Identity refs are not reused on queued extensions.</p>
+            <p v-if="useIdentityRefs" class="text-xs text-zinc-500">Picture 1 is the start still and stays Picture 1 on every shot, including queued extensions. Extra stills map to Picture 2–5. Empty extra slots reuse the start still.</p>
           </div>
         </div>
 

server/api/generate.post.ts

@@ -309,17 +309,17 @@ async function queueMiniMax(
 
   const done = watchComfyJob(job, { persist: params.persist })
   job.status = 'uploading'
-  const uploading = (job.library?.chainIndex || 0) > 0
-    ? 'Uploading last frame to ComfyUI...'
-    : 'Uploading image to ComfyUI...'
+  const chainIndex = job.library?.chainIndex || 0
+  const uploading = params.useIdentityRefs
+    ? (chainIndex > 0 ? 'Uploading identity stills for next shot...' : 'Uploading image to ComfyUI...')
+    : (chainIndex > 0 ? 'Uploading last frame to ComfyUI...' : 'Uploading image to ComfyUI...')
   const queueing = (job.library?.chainIndex || 0) > 0
     ? 'Queueing extension on MiniMax H3...'
     : 'Queueing MiniMax H3 job...'
   emitChainJob(job, { type: 'status', message: uploading, progress: 4 })
   const uploaded = await uploadImage(params.image, job.id)
   const referenceNames: string[] = ['', '', '', '']
-  const chainIndex = job.library?.chainIndex || 0
-  if (chainIndex === 0) {
+  if (params.useIdentityRefs) {
     for (const [index, ref] of (params.referenceImages || []).entries()) {
       if (!ref?.data?.length) continue
       const next = await uploadImage({
@@ -353,8 +353,8 @@ async function queueMiniMax(
     sound: params.sound,
     workflow: params.workflow,
     duration: params.duration,
-    useIdentityRefs: chainIndex === 0 && params.useIdentityRefs,
-    referenceImageNames: chainIndex === 0 ? referenceNames : []
+    useIdentityRefs: params.useIdentityRefs,
+    referenceImageNames: params.useIdentityRefs ? referenceNames : []
   })
 
   const queued = await queuePrompt(graph, job.clientId)
@@ -466,7 +466,9 @@ async function runGeneration(
     await ensureComfyReady(ready)
     await queueMiniMax(job, {
       prompt: ext.prompt,
-      image: { filename: 'last_frame.png', data: frame, type: 'image/png' },
+      image: params.useIdentityRefs
+        ? params.image
+        : { filename: 'last_frame.png', data: frame, type: 'image/png' },
       width: params.width,
       height: params.height,
       steps: params.steps,
@@ -481,8 +483,8 @@ async function runGeneration(
       persist: isLast,
       workflow: params.workflow,
       duration: ext.duration,
-      useIdentityRefs: false,
-      referenceImages: []
+      useIdentityRefs: params.useIdentityRefs,
+      referenceImages: params.useIdentityRefs ? params.referenceImages : []
     })
 
     if (!isLast) {

utils/identityPrompt.ts

@@ -13,11 +13,11 @@ export function identityBoilerplate(extraRefs: number | number[] = 0) {
     : extras.length === 1
       ? ` <Picture ${extras[0]}> is a secondary reference angle for profile details.`
       : ` ${extras.map(n => `<Picture ${n}>`).join(' and ')} are secondary reference angles for profile features only.`
-  return `<Picture 1> is the exact facial identity, hairstyle, body proportions and clothing of the main character. Preserve this identity with high fidelity even when the subject temporarily leaves the frame and returns.${secondary}`
+  return `<Picture 1> is the primary identity anchor defining exact facial geometry, hairstyle, and body proportions.${secondary} Strictly maintain the structural proportions of <Picture 1> with high fidelity when moving.`
 }
 
 export function looksLikeIdentityPrompt(text: string) {
-  return /primary identity anchor|exact facial identity/i.test(String(text || ''))
+  return /^<Picture 1> is the (primary identity anchor|exact facial identity)/i.test(String(text || '').trim())
 }
 
 export function buildIdentityPrompt(actionPrompt: string, extraRefs: number | number[] = 0) {