vibe-aigc 0.7.0__py3-none-any.whl → 0.7.1__py3-none-any.whl

vibe_aigc/tools_comfyui.py

@@ -232,8 +232,9 @@ class VideoGenerationTool(BaseTool):
     """
     Video generation tool using local ComfyUI.
     
-    Uses the I2V pipeline: FLUX → Wan I2V animation.
-    MetaPlanner can use this for motion content.
+    Supports two modes:
+    - With image_url: Uses I2V (Image-to-Video) to animate existing image
+    - Without image_url: Uses T2V (Text-to-Video) to generate from scratch
     """
     
     def __init__(self, comfyui_url: str = "http://127.0.0.1:8188"):
@@ -244,13 +245,14 @@ class VideoGenerationTool(BaseTool):
     def spec(self) -> ToolSpec:
         return ToolSpec(
             name="video_generation",
-            description="Generate videos using local AI models (FLUX + Wan I2V)",
+            description="Generate videos using local AI models. Pass image_url to animate an existing image (I2V), or just prompt for text-to-video.",
             category=ToolCategory.VIDEO,
             input_schema={
                 "type": "object",
                 "required": ["prompt"],
                 "properties": {
-                    "prompt": {"type": "string", "description": "Video description"},
+                    "prompt": {"type": "string", "description": "Video/motion description"},
+                    "image_url": {"type": "string", "description": "Source image URL to animate (enables I2V mode)"},
                     "negative_prompt": {"type": "string", "description": "What to avoid"},
                     "style": {"type": "string", "description": "Visual style hints"},
                     "motion": {"type": "string", "description": "Motion description"},
@@ -269,7 +271,7 @@ class VideoGenerationTool(BaseTool):
             },
             examples=[
                 {
-                    "input": {"prompt": "samurai walking through rain", "frames": 33},
+                    "input": {"prompt": "gentle breeze, swaying flowers", "image_url": "http://.../image.png"},
                     "output": {"video_url": "http://...", "duration_seconds": 2.0}
                 }
             ]
@@ -299,6 +301,7 @@ class VideoGenerationTool(BaseTool):
             prompt = inputs.get("prompt", "")
             style = inputs.get("style", "")
             motion = inputs.get("motion", "")
+            image_url = inputs.get("image_url")
             
             # Incorporate knowledge base hints
             if context and "technical_specs" in context:
@@ -316,15 +319,22 @@ class VideoGenerationTool(BaseTool):
             frames = inputs.get("frames", 33)
             fps = inputs.get("fps", 16)
             
+            # Choose capability based on whether we have a source image
+            if image_url:
+                capability = Capability.IMAGE_TO_VIDEO
+            else:
+                capability = Capability.TEXT_TO_VIDEO
+            
             result = await backend.generate(GenerationRequest(
                 prompt=prompt,
-                capability=Capability.TEXT_TO_VIDEO,
+                capability=capability,
                 negative_prompt=inputs.get("negative_prompt", ""),
                 width=inputs.get("width", 832),
                 height=inputs.get("height", 480),
                 frames=frames,
                 steps=20,
-                cfg=5.0
+                cfg=5.0,
+                image_url=image_url  # Pass source image for I2V
             ))
             
             if result.success:

vibe_aigc/vibe_backend.py

@@ -41,6 +41,7 @@ class GenerationRequest:
     steps: int = 20
     cfg: float = 7.0
     seed: Optional[int] = None
+    image_url: Optional[str] = None  # Source image for I2V/img2img
 
 
 @dataclass
@@ -150,10 +151,18 @@ class VibeBackend:
             import random
             request.seed = random.randint(0, 2**32 - 1)
         
-        # Special handling for TEXT_TO_VIDEO: use I2V pipeline
+        # Special handling for video capabilities
         if request.capability == Capability.TEXT_TO_VIDEO:
             return await self._generate_video_via_i2v(request)
         
+        # IMAGE_TO_VIDEO: animate a provided image
+        if request.capability == Capability.IMAGE_TO_VIDEO:
+            if request.image_url:
+                return await self._animate_image(request)
+            else:
+                # No image provided, fall back to T2V pipeline
+                return await self._generate_video_via_i2v(request)
+        
         # Try to get workflow
         workflow = await self._get_workflow(request)
         if not workflow:
@@ -467,6 +476,54 @@ class VibeBackend:
         print(f"    Video: {video_result.output_path}")
         return video_result
     
+    async def _animate_image(self, request: GenerationRequest) -> GenerationResult:
+        """Animate an existing image with I2V.
+        
+        Uses provided image_url instead of generating a new base image.
+        """
+        print(f"\n[I2V] Animating provided image...")
+        print(f"    Source: {request.image_url}")
+        
+        # Download the source image
+        async with aiohttp.ClientSession() as session:
+            async with session.get(request.image_url) as resp:
+                if resp.status != 200:
+                    return GenerationResult(
+                        success=False,
+                        error=f"Failed to download image: HTTP {resp.status}"
+                    )
+                image_data = await resp.read()
+            
+            # Upload to ComfyUI
+            form = aiohttp.FormData()
+            form.add_field('image', image_data, filename='input.png', content_type='image/png')
+            
+            async with session.post(f"{self.url}/upload/image", data=form) as resp:
+                upload_result = await resp.json()
+                uploaded_name = upload_result.get("name", "input.png")
+                print(f"    Uploaded: {uploaded_name}")
+        
+        # Create I2V workflow
+        i2v_workflow = self._create_wan_i2v_workflow(
+            uploaded_image=uploaded_name,
+            prompt=request.prompt,
+            negative=request.negative_prompt,
+            width=request.width,
+            height=request.height,
+            frames=request.frames,
+            seed=request.seed
+        )
+        
+        video_result = await self._execute_workflow(i2v_workflow)
+        if not video_result.success:
+            return GenerationResult(
+                success=False,
+                error=f"Animation failed: {video_result.error}"
+            )
+        
+        print(f"    Video: {video_result.output_path}")
+        return video_result
+    
     def _create_flux_image_workflow(
         self, prompt: str, negative: str, width: int, height: int, seed: int
     ) -> Dict[str, Any]:

{vibe_aigc-0.7.0.dist-info → vibe_aigc-0.7.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vibe-aigc
-Version: 0.7.0
+Version: 0.7.1
 Summary: A New Paradigm for Content Generation via Agentic Orchestration
 Author: Vibe AIGC Contributors
 License-Expression: MIT

{vibe_aigc-0.7.0.dist-info → vibe_aigc-0.7.1.dist-info}/RECORD

@@ -19,12 +19,12 @@ vibe_aigc/pipeline.py,sha256=sWoZq0mu9_ttLbzIJ0gZL8_1X9lH3iyK0mfi5VWufrQ,19032
 vibe_aigc/planner.py,sha256=TfndK417ZsoomDkDGGBx1SXDUUu8e_h5c4zELLIsNXQ,38448
 vibe_aigc/tools.py,sha256=crpUqqHEGQ7H-p5qfhYSCYgrIeR4Gx_2heEV_t9m6ak,17127
 vibe_aigc/tools_audio.py,sha256=cup4-Zo-cRTqnGAoWYmgQBK_tLl6sac7s9uSRyeZKNQ,25374
-vibe_aigc/tools_comfyui.py,sha256=XtPwAbQ6NcfgXjQokfPSc8aklMYFVRl51epwXZcWgjw,37924
+vibe_aigc/tools_comfyui.py,sha256=hj-cP2k28KmL3DvsVIGVlvgJssavR-9vihdDc7W-eHg,38550
 vibe_aigc/tools_multimodal.py,sha256=asSJJqF0hrD9uNiYpuieVY-lbgEXjbK3UjT20nX2Lig,20405
 vibe_aigc/tools_utility.py,sha256=oDWursVXUhbxBdKFT0MmHS2Up9d8X0Ff16Cb0JREBGc,33754
 vibe_aigc/tools_video.py,sha256=zTaVMsz6E_R74CxnJ7iFfbB1othxi6EzokD1X6a-DTs,27800
 vibe_aigc/tools_vision.py,sha256=eItQ8feTsdbkVOlTZ67317IYu4t6OWHKmsO2w7sNp4A,43795
-vibe_aigc/vibe_backend.py,sha256=WStTXrtL73vSbqRA9hA72zcrV8Xt9gpZdQ0lk-M3lLw,24475
+vibe_aigc/vibe_backend.py,sha256=YoBJrC-LcoVjHoDerxpeodNs9LvErr85KaJ6Q0fc5yA,26819
 vibe_aigc/video.py,sha256=0fg8RUpEsaJqDskAPiGP8yuyQDVCUvIy-uLScq_BOwg,14111
 vibe_aigc/visualization.py,sha256=jDs2f1vj4k8ZnJTA_niKLBH2NMahTgWneiADlNmW24s,7143
 vibe_aigc/vlm_feedback.py,sha256=AL_QJKLs80tsUcYhd1FwddI783j-ADVg_2xbK7vPOj4,19288
@@ -34,9 +34,9 @@ vibe_aigc/workflow_executor.py,sha256=mfYLOTfPmI7Upooxy07nPmlbZ-HZAfC18IaNW80G31
 vibe_aigc/workflow_registry.py,sha256=Z6gB1cA366LXqHcfqBF1od_8ySxAOt5RpKKaaZPqqUo,22359
 vibe_aigc/workflow_strategies.py,sha256=i_qqUrn-2F6lT9dNyFdTdy0NzE8ZnRNxAMl6zrOAtD8,26148
 vibe_aigc/workflows.py,sha256=uk7RjNVow6eimEdqfVQFDtLgHSkg0LUjSoa2N7C47u0,13886
-vibe_aigc-0.7.0.dist-info/licenses/LICENSE,sha256=Ir4dCTvOsbfoiOh9vYbhIKDH59S7J6qhJYZmHHICoKY,1079
-vibe_aigc-0.7.0.dist-info/METADATA,sha256=_TvMtrNMqPC5i9SeURKvmFcKleahr2HRA6x1K_nIbcQ,7391
-vibe_aigc-0.7.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-vibe_aigc-0.7.0.dist-info/entry_points.txt,sha256=2htp4yXJMvCAQXTB39XWWwbBPP3MYUYXsqlwMeQsd7o,49
-vibe_aigc-0.7.0.dist-info/top_level.txt,sha256=Cpjz8X0WEhnhaigqxmsZSl9VxduaDspj7WuVUGGLeao,10
-vibe_aigc-0.7.0.dist-info/RECORD,,
+vibe_aigc-0.7.1.dist-info/licenses/LICENSE,sha256=Ir4dCTvOsbfoiOh9vYbhIKDH59S7J6qhJYZmHHICoKY,1079
+vibe_aigc-0.7.1.dist-info/METADATA,sha256=_TbKMeACL5FDeKXRQKhdkBkzdBOLm7L_PPf-3tU19b4,7391
+vibe_aigc-0.7.1.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+vibe_aigc-0.7.1.dist-info/entry_points.txt,sha256=2htp4yXJMvCAQXTB39XWWwbBPP3MYUYXsqlwMeQsd7o,49
+vibe_aigc-0.7.1.dist-info/top_level.txt,sha256=Cpjz8X0WEhnhaigqxmsZSl9VxduaDspj7WuVUGGLeao,10
+vibe_aigc-0.7.1.dist-info/RECORD,,