Update app.py

app.py

@@ -70,189 +70,76 @@ def generate_video(
     # Create a unique ID for this generation
     generation_id = uuid.uuid4().hex[:8]
     print(f"Generation ID: {generation_id}")
-
-    # Initial parameters dictionary - we'll customize it for each provider
-    base_parameters = {
-        "negative_prompt": negative_prompt,
-        "num_inference_steps": num_inference_steps,
-        "guidance_scale": guidance_scale,
-    }
-    
-    # Add seed if specified
-    if seed is not None and seed != -1:
-        base_parameters["seed"] = seed
     
-    # Provider-specific parameter handling
-    if provider == "hf-inference":
-        print("Using Hugging Face Inference API, adapting parameters...")
-        # HF Inference API supports these parameters
-        parameters = {
-            "negative_prompt": negative_prompt,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-            "num_frames": num_frames,
-            # HF Inference does NOT support fps
-            # "fps": fps,  
-            "width": width,
-            "height": height,
+    # Define supported parameters for each provider
+    provider_param_support = {
+        "hf-inference": {
+            "supported": ["prompt", "model", "negative_prompt", "num_frames", "num_inference_steps", "guidance_scale", "seed"],
+            "extra_info": "HF Inference doesn't support 'fps', 'width', 'height', or 'motion_bucket_id' parameters"
+        },
+        "fal-ai": {
+            "supported": ["prompt", "model", "negative_prompt", "num_frames", "num_inference_steps", "guidance_scale", "seed"],
+            "extra_info": "Fal-AI doesn't support 'fps', 'width', 'height', or 'motion_bucket_id' parameters"
+        },
+        "novita": {
+            "supported": ["prompt", "model", "negative_prompt", "num_frames", "num_inference_steps", "guidance_scale", "seed", "fps", "width", "height"],
+            "extra_info": "Novita may not support 'motion_bucket_id' parameter"
+        },
+        "replicate": {
+            "supported": ["prompt", "model", "negative_prompt", "num_frames", "num_inference_steps", "guidance_scale", "seed", "fps", "width", "height"],
+            "extra_info": "Replicate parameters vary by specific model"
         }
-        
-        # Add motion_bucket_id for SVD models if applicable
-        if "stable-video-diffusion" in model_to_use:
-            parameters["motion_bucket_id"] = motion_bucket_id
-            
-        # Add seed if specified
-        if seed is not None and seed != -1:
-            parameters["seed"] = seed
+    }
     
-    elif provider == "fal-ai":
-        print("Using FalAI provider, adapting parameters...")
-        # According to Fal-AI API specification, only these parameters are supported
-        parameters = {
-            "negative_prompt": negative_prompt,
-            "num_frames": num_frames,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-        }
-        
-        # Add seed if specified
-        if seed is not None and seed != -1:
-            parameters["seed"] = seed
-        
-        # Note: width and height are not supported by Fal-AI's text_to_video API
-        print("Note: width and height parameters are not supported by Fal-AI and will be ignored")
+    # Get supported parameters for the current provider
+    supported_params = provider_param_support.get(provider, {}).get("supported", [])
+    provider_info = provider_param_support.get(provider, {}).get("extra_info", "No specific information available")
     
-    elif provider == "novita":
-        print("Using Novita provider, adapting parameters...")
-        # Based on documentation, Novita uses specific parameters
-        parameters = {
-            "negative_prompt": negative_prompt,
-            "num_frames": num_frames,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-            "width": width,
-            "height": height
-        }
-        
-        # Add seed if specified
-        if seed is not None and seed != -1:
-            parameters["seed"] = seed
+    print(f"Provider info: {provider_info}")
+    print(f"Supported parameters: {supported_params}")
     
-    elif provider == "replicate":
-        print("Using Replicate provider, adapting parameters...")
-        parameters = {
-            "negative_prompt": negative_prompt,
-            "num_frames": num_frames,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-            "width": width,
-            "height": height
-        }
-        
-        # Add seed if specified
-        if seed is not None and seed != -1:
-            parameters["seed"] = seed
-        
-        # Replicate supports fps in some models
-        if "stable-video-diffusion" in model_to_use:
-            parameters["fps"] = fps
+    # Create a parameters dictionary with only supported parameters
+    parameters = {}
     
-    else:
-        # Default parameters for any other provider
-        print(f"Using generic parameters for provider: {provider}")
-        parameters = base_parameters
+    if "negative_prompt" in supported_params:
+        parameters["negative_prompt"] = negative_prompt
+    
+    if "num_frames" in supported_params:
         parameters["num_frames"] = num_frames
+    
+    if "num_inference_steps" in supported_params:
+        parameters["num_inference_steps"] = num_inference_steps
+    
+    if "guidance_scale" in supported_params:
+        parameters["guidance_scale"] = guidance_scale
+    
+    if "seed" in supported_params and seed is not None:
+        parameters["seed"] = seed
+    
+    if "fps" in supported_params:
+        parameters["fps"] = fps
+    
+    if "width" in supported_params:
         parameters["width"] = width
+    
+    if "height" in supported_params:
         parameters["height"] = height
-        "height": height,
-        "num_inference_steps": num_inference_steps,
-        "guidance_scale": guidance_scale,
-    }
     
-    # Add motion_bucket_id if applicable (depends on the model)
-    if motion_bucket_id is not None:
+    if "motion_bucket_id" in supported_params:
         parameters["motion_bucket_id"] = motion_bucket_id
-    
-    # Add seed if specified
-    if seed is not None:
-        parameters["seed"] = seed
 
-    # For FalAI provider - may need specific formatting
-    if provider == "fal-ai":
-        print("Using FalAI provider, adapting parameters...")
-        # According to Fal-AI API specification, only these parameters are supported
-        parameters = {
-            "negative_prompt": negative_prompt,
-            "num_frames": num_frames,
-            "num_inference_steps": num_inference_steps,
-            "guidance_scale": guidance_scale,
-        }
-        
-        # Add seed if specified
-        if seed is not None and seed != -1:
-            parameters["seed"] = seed
-        
-        # Note: width and height are not supported by Fal-AI's text_to_video API
-        print("Note: width and height parameters are not supported by Fal-AI and will be ignored")
+    # Now that we have a clean parameter set, handle provider-specific logic
+    print(f"Final parameters for {provider}: {parameters}")
     
-    # For Novita provider - may need specific formatting
-    if provider == "novita":
-        print("Using Novita provider, adapting parameters...")
-        # Based on documentation, Novita uses text_to_video method
-        try:
-            # For Novita, we create a dedicated parameters object
-            novita_params = {
-                "negative_prompt": negative_prompt,
-                "num_frames": num_frames,
-                "fps": fps,
-                "width": width,
-                "height": height,
-                "num_inference_steps": num_inference_steps,
-                "guidance_scale": guidance_scale
-            }
-            
-            # Add seed if specified
-            if seed is not None:
-                novita_params["seed"] = seed
-                
-            # For Novita, we use a different method from the InferenceClient
-            video_data = client.text_to_video(
-                prompt=prompt,
-                model=model_to_use,
-                **novita_params
-            )
-            
-            # Save the video to a temporary file
-            temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
-            temp_file.write(video_data)
-            video_path = temp_file.name
-            temp_file.close()
-            
-            print(f"Video saved to temporary file: {video_path}")
-            return video_path
-            
-        except Exception as e:
-            print(f"Error during Novita video generation: {e}")
-            return f"Error: {str(e)}"
-    
-    # For Replicate provider - may need specific formatting
+    # For Replicate provider - uses post method
     if provider == "replicate":
-        print("Using Replicate provider, adapting parameters...")
-        # Replicate might use different parameter formats
+        print("Using Replicate provider, using post method...")
         try:
-            # For Replicate, we use their specific method structure
             response = client.post(
                 model=model_to_use,
                 input={
                     "prompt": prompt,
-                    "negative_prompt": negative_prompt,
-                    "num_frames": num_frames,
-                    "fps": fps,
-                    "width": width,
-                    "height": height,
-                    "num_inference_steps": num_inference_steps,
-                    "guidance_scale": guidance_scale,
-                    "seed": seed if seed is not None else 0,
+                    **parameters
                 },
             )
             
@@ -268,16 +155,11 @@ def generate_video(
             print(f"Error during Replicate video generation: {e}")
             return f"Error: {str(e)}"
     
-    # General approach for other providers
+    # For all other providers, use the standard text_to_video method
     try:
         print(f"Sending request to {provider} provider with model {model_to_use}.")
-        print(f"Parameters: {parameters}")
         
-        # Make sure prompt is not in parameters to avoid duplicate parameters
-        if "prompt" in parameters:
-            del parameters["prompt"]
-            
-        # Use the text_to_video method of the InferenceClient
+        # Use the text_to_video method of the InferenceClient with only supported parameters
         video_data = client.text_to_video(
             prompt=prompt,
             model=model_to_use,
@@ -369,40 +251,6 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
                     lines=2
                 )
                 
-                with gr.Row():
-                    num_inference_steps = gr.Slider(
-                        minimum=1,
-                        maximum=100,
-                        value=25,
-                        step=1,
-                        label="Inference Steps"
-                    )
-                    
-                    guidance_scale = gr.Slider(
-                        minimum=1.0,
-                        maximum=20.0,
-                        value=7.5,
-                        step=0.5,
-                        label="Guidance Scale"
-                    )
-                
-                with gr.Row():
-                    motion_bucket_id = gr.Slider(
-                        minimum=1,
-                        maximum=255,
-                        value=127,
-                        step=1,
-                        label="Motion Bucket ID (for SVD models)"
-                    )
-                    
-                    seed = gr.Slider(
-                        minimum=-1,
-                        maximum=2147483647,
-                        value=-1,
-                        step=1,
-                        label="Seed (-1 for random)"
-                    )
-                
                 with gr.Row():
                     width = gr.Slider(
                         minimum=256,