Major Story Reels update: Direct script mode, sticky animation style, improved scene continuity prompts

modules/story_reels/router.py

@@ -34,23 +34,22 @@ router = APIRouter()
     response_model=GenerateVideoResponse,
     status_code=201,
     summary="Create a new story reel",
-    description="Create a new AI-generated story video. Returns video_id to track progress."
+    description="Create a story video from script. Returns video_id to track progress."
 )
 async def create_story_reel(request: GenerateVideoRequest):
     """
-    Create a new story reel from topic.
+    Create a new story reel from script.
     
-    - AI generates script from topic
     - Converts script to speech (TTS)
     - Generates captions (Whisper)
     - Creates AI images (NVIDIA/Cloudflare)
     - Composes final video (MoviePy)
     """
     try:
-        logger.info(f"Creating story reel for topic: {request.topic}")
+        logger.info(f"Creating story reel from direct script ({len(request.script)} chars)")
         
         video_id = story_creator.add_to_queue(
-            topic=request.topic,
+            script=request.script,
             image_style=request.image_style.value,
             voice=request.voice
         )

modules/story_reels/schemas.py

@@ -14,6 +14,7 @@ class StyleEnum(str, Enum):
     cartoon = "cartoon"
     realistic = "realistic"
     watercolor = "watercolor"
+    sticky_animation = "sticky animation"
 
 
 class CameraEnum(str, Enum):
@@ -80,8 +81,8 @@ class GeneratedScene(BaseModel):
 # ===================
 
 class GenerateVideoRequest(BaseModel):
-    """Main video generation request - simplified"""
-    topic: str = Field(..., description="Video topic/title (include any specific requirements here)")
+    """Main video generation request - direct script mode"""
+    script: str = Field(..., description="Voice script text (will be converted to TTS directly)")
     image_style: StyleEnum = Field(StyleEnum.semi_realistic, description="Image generation style")
     voice: str = Field("af_heart", description="TTS voice")
 

modules/story_reels/services/script_generator.py

@@ -151,16 +151,31 @@ Your task: Generate detailed image prompts for each 2-second scene of a story vi
 CONTEXT:
 - Full story script is provided so you understand the narrative
 - Each 2-second chunk needs a visual prompt
-- Character profile (if provided) must be consistent in EVERY prompt
-- Images should tell the story visually
+- Images will play in SEQUENCE to tell a story
+- All images MUST look like they belong to the SAME VIDEO
 
-RULES FOR PROMPTS:
-1. Be detailed and specific (50-100 words each)
-2. Include: scene description, character pose/action, camera angle, lighting, mood
-3. Add style keywords at the end (semi-realistic, detailed, high quality)
-4. DO NOT include text/dialogue in prompts
-5. Keep character appearance CONSISTENT across all prompts
-6. Use cinematographic language (close-up, wide shot, etc.)
+CRITICAL RULES FOR CONSISTENCY:
+1. SAME STYLE: Every prompt MUST start with the exact style name (e.g., "semi-realistic style", "anime style", "sticky animation style")
+2. SAME CHARACTER: If a character is described, use IDENTICAL description in EVERY prompt (same clothes, hair, face features)
+3. SCENE CONTINUITY: Each scene should logically follow the previous one
+   - Example: Scene 1 "boy picking up bag" → Scene 2 "boy walking with bag on shoulder" → Scene 3 "boy approaching school gate"
+4. CONSISTENT LIGHTING: Use same lighting style across all scenes
+5. CONSISTENT COLOR PALETTE: Maintain similar color tones
+
+PROMPT STRUCTURE:
+1. [STYLE] - Always start with style (e.g., "semi-realistic style artwork")
+2. [CHARACTER] - Describe the character with exact same details every time
+3. [ACTION] - What's happening in THIS specific 2-second moment
+4. [ENVIRONMENT] - Where is this taking place
+5. [CAMERA] - Camera angle (close-up, medium shot, wide shot)
+6. [LIGHTING & MOOD] - Lighting and emotional atmosphere
+7. [QUALITY TAGS] - high quality, detailed, cinematic, 8k
+
+CONTINUITY TIPS:
+- If character was sitting, show transition to standing (not jumping to running)
+- Keep background elements consistent (same room, same street)
+- Props should persist (if bag appeared, keep showing it)
+- Time progression should be logical
 
 OUTPUT FORMAT:
 Return ONLY valid JSON array, no markdown, no explanation:

modules/story_reels/services/story_creator.py

@@ -58,13 +58,16 @@ class StoryCreator:
         
     def add_to_queue(
         self,
-        topic: str,
+        script: str,
         image_style: str = "semi-realistic",
         voice: str = "af_heart"
     ) -> str:
         """
         Add story to generation queue.
         
+        Args:
+            script: Voice script text (will be converted to TTS directly)
+        
         Returns:
             job_id for tracking
         """
@@ -72,7 +75,7 @@ class StoryCreator:
         
         job = {
             "id": job_id,
-            "topic": topic,
+            "script": script,
             "image_style": image_style,
             "voice": voice,
             "status": JobStatus.queued,
@@ -195,19 +198,14 @@ class StoryCreator:
         
         try:
             # ====================
-            # Step 0: Generate Script from Topic
+            # Step 1: Use Script Directly (No AI Generation)
             # ====================
-            logger.info(f"[{job_id}] Generating script from topic using AI...")
+            script = job["script"]
+            logger.info(f"[{job_id}] Using direct script: {len(script)} chars")
             job["progress"] = 5
             
-            script = self.script_gen.generate_script(
-                topic=job["topic"],
-                max_chars=1000
-            )
-            logger.info(f"[{job_id}] Generated script: {len(script)} chars")
-            
             # ====================
-            # Step 1: Generate TTS
+            # Step 2: Generate TTS
             # ====================
             job["status"] = JobStatus.generating_audio
             job["progress"] = 10

static/index.html

@@ -285,12 +285,12 @@
 
                 <form id="storyForm">
                     <div class="form-group">
-                        <label>Topic / Idea *</label>
-                        <textarea id="storyTopic" rows="3"
-                            placeholder="e.g., A 16-year-old boy named Rafi going to school on his first day. He has short black hair, light brown skin, wearing casual clothes."
+                        <label>Voice Script *</label>
+                        <textarea id="storyScript" rows="4"
+                            placeholder="e.g., Have you ever wondered why some people seem to attract success effortlessly? The secret lies in their mindset. When you believe in yourself, opportunities start flowing toward you..."
                             required></textarea>
-                        <small style="color: var(--text-secondary);">Include character details, setting, mood -
-                            everything in the topic!</small>
+                        <small style="color: var(--text-secondary);">Enter the exact script you want converted to voice.
+                            This will be spoken directly.</small>
                     </div>
 
                     <div class="form-row">
@@ -302,6 +302,7 @@
                                 <option value="cartoon">Cartoon</option>
                                 <option value="realistic">Realistic</option>
                                 <option value="watercolor">Watercolor</option>
+                                <option value="sticky animation">Sticky Animation</option>
                             </select>
                         </div>
                         <div class="form-group">
@@ -479,7 +480,7 @@
             status.classList.remove('hidden');
 
             const data = {
-                topic: document.getElementById('storyTopic').value,
+                script: document.getElementById('storyScript').value,
                 image_style: document.getElementById('storyStyle').value,
                 voice: document.getElementById('storyVoice').value
             };