show ffmpg command in case of error

app.py

@@ -12,24 +12,13 @@ import tempfile
 import shlex
 import shutil
 
-# Supported models configuration
-MODELS = {
-    "deepseek-ai/DeepSeek-V3": {
-        "base_url": "https://api.deepseek.com/v1",
-        "env_key": "DEEPSEEK_API_KEY",
-    },
-}
-
-# Initialize client with first available model
-client = OpenAI(
-    base_url=next(iter(MODELS.values()))["base_url"],
-    api_key=os.environ[next(iter(MODELS.values()))["env_key"]],
-)
+HF_API_KEY = os.environ["HF_TOKEN"]
+
+client = OpenAI(base_url="https://api-inference.huggingface.co/v1/", api_key=HF_API_KEY)
 
 allowed_medias = [
     ".png",
     ".jpg",
-    ".webp",
     ".jpeg",
     ".tiff",
     ".bmp",
@@ -95,7 +84,7 @@ def get_files_infos(files):
     return results
 
 
-def get_completion(prompt, files_info, top_p, temperature, model_choice):
+def get_completion(prompt, files_info, top_p, temperature):
     # Create table header
     files_info_string = "| Type | Name | Dimensions | Duration | Audio Channels |\n"
     files_info_string += "|------|------|------------|-----------|--------|\n"
@@ -128,7 +117,6 @@ You are given:
 Your objective is to generate the SIMPLEST POSSIBLE single ffmpeg command to create the requested video.
 
 Key requirements:
-    - First, think step-by-step about what the user is asking for and reformulate it into a clear technical specification
     - Use the absolute minimum number of ffmpeg options needed
     - Avoid complex filter chains or filter_complex if possible
     - Prefer simple concatenation, scaling, and basic filters
@@ -144,19 +132,15 @@ Remember: Simpler is better. Only use advanced ffmpeg features if absolutely nec
         },
         {
             "role": "user",
-            "content": f"""Always output the media as video/mp4 and output file with "output.mp4".
-The current assets and objective follow.
+            "content": f"""Always output the media as video/mp4 and output file with "output.mp4". Provide only the shell command without any explanations.
+The current assets and objective follow. Reply with the FFMPEG command:
 
 AVAILABLE ASSETS LIST:
 
 {files_info_string}
 
 OBJECTIVE: {prompt} and output at "output.mp4"
-
-First, think step-by-step about what I'm asking for and reformulate it into a clear technical specification.
-Then provide the FFMPEG command that will accomplish this task.
-
-YOUR RESPONSE:
+YOUR FFMPEG COMMAND:
          """,
         },
     ]
@@ -168,16 +152,8 @@ YOUR RESPONSE:
             print(msg["content"])
         print("=====================\n")
 
-        if model_choice not in MODELS:
-            raise ValueError(f"Model {model_choice} is not supported")
-
-        model_config = MODELS[model_choice]
-        client.base_url = model_config["base_url"]
-        client.api_key = os.environ[model_config["env_key"]]
-        model = "deepseek-chat" if "deepseek" in model_choice.lower() else model_choice
-
         completion = client.chat.completions.create(
-            model=model,
+            model="Qwen/Qwen2.5-Coder-32B-Instruct",
             messages=messages,
             temperature=temperature,
             top_p=top_p,
@@ -189,23 +165,13 @@ YOUR RESPONSE:
             # Find content between ```sh or ```bash and the next ```
             import re
 
-            command_match = re.search(r"```(?:sh|bash)?\n(.*?)\n```", content, re.DOTALL)
-            if command_match:
-                command = command_match.group(1).strip()
-            else:
-                # Try to find a line that starts with ffmpeg
-                ffmpeg_lines = [line.strip() for line in content.split('\n') if line.strip().startswith('ffmpeg')]
-                if ffmpeg_lines:
-                    command = ffmpeg_lines[0]
-                else:
-                    command = content.replace("\n", "")
-        else:
-            # Try to find a line that starts with ffmpeg
-            ffmpeg_lines = [line.strip() for line in content.split('\n') if line.strip().startswith('ffmpeg')]
-            if ffmpeg_lines:
-                command = ffmpeg_lines[0]
+            command = re.search(r"```(?:sh|bash)?\n(.*?)\n```", content, re.DOTALL)
+            if command:
+                command = command.group(1).strip()
             else:
                 command = content.replace("\n", "")
+        else:
+            command = content.replace("\n", "")
 
         # remove output.mp4 with the actual output file path
         command = command.replace("output.mp4", "")
@@ -215,13 +181,7 @@ YOUR RESPONSE:
         raise Exception("API Error")
 
 
-def update(
-    files,
-    prompt,
-    top_p=1,
-    temperature=1,
-    model_choice="deepseek-ai/DeepSeek-V3",
-):
+def update(files, prompt, top_p=1, temperature=1):
     if prompt == "":
         raise gr.Error("Please enter a prompt.")
 
@@ -233,16 +193,14 @@ def update(
                 raise gr.Error(
                     "Please make sure all videos are less than 2 minute long."
                 )
-        if file_info["size"] > 100000000:
-            raise gr.Error("Please make sure all files are less than 100MB in size.")
+        if file_info["size"] > 10000000:
+            raise gr.Error("Please make sure all files are less than 10MB in size.")
 
     attempts = 0
     while attempts < 2:
         print("ATTEMPT", attempts)
         try:
-            command_string = get_completion(
-                prompt, files_info, top_p, temperature, model_choice
-            )
+            command_string = get_completion(prompt, files_info, top_p, temperature)
             print(
                 f"""///PROMTP {prompt} \n\n/// START OF COMMAND ///:\n\n{command_string}\n\n/// END OF COMMAND ///\n\n"""
             )
@@ -281,9 +239,7 @@ def update(
                 f"\n=== EXECUTING FFMPEG COMMAND ===\nffmpeg {' '.join(final_command[1:])}\n"
             )
             subprocess.run(final_command, cwd=temp_dir)
-            # Extract just the command for display
-            command_for_display = f"ffmpeg {' '.join(args[1:])} -y output.mp4"
-            generated_command = f"### Generated Command\n```bash\n{command_for_display}\n```"
+            generated_command = f"### Generated Command\n```bash\nffmpeg {' '.join(args[1:])} -y output.mp4\n```"
             return output_file_path, gr.update(value=generated_command)
         except Exception as e:
             attempts += 1
@@ -296,7 +252,7 @@ with gr.Blocks() as demo:
     gr.Markdown(
         """
             # 🏞 AI Video Composer
-            Compose new videos from your assets using natural language. Add video, image and audio assets and let [Qwen2.5-Coder](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) or [DeepSeek-V3](https://huggingface.co/deepseek-ai/DeepSeek-V3-Base) generate a new video for you (using FFMPEG).
+            Compose new videos from your assets using natural language. Add video, image and audio assets and let [Qwen2.5-Coder](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) generate a new video for you (using FFMPEG).
         """,
         elem_id="header",
     )
@@ -308,16 +264,11 @@ with gr.Blocks() as demo:
                 file_types=allowed_medias,
             )
             user_prompt = gr.Textbox(
-                placeholder="eg: Remove the 3 first seconds of the video",
+                placeholder="I want to convert to a gif under 15mb",
                 label="Instructions",
             )
             btn = gr.Button("Run")
             with gr.Accordion("Parameters", open=False):
-                model_choice = gr.Radio(
-                    choices=list(MODELS.keys()),
-                    value=list(MODELS.keys())[0],
-                    label="Model",
-                )
                 top_p = gr.Slider(
                     minimum=-0,
                     maximum=1.0,
@@ -342,7 +293,7 @@ with gr.Blocks() as demo:
 
         btn.click(
             fn=update,
-            inputs=[user_files, user_prompt, top_p, temperature, model_choice],
+            inputs=[user_files, user_prompt, top_p, temperature],
             outputs=[generated_video, generated_command],
         )
     with gr.Row():
@@ -350,35 +301,40 @@ with gr.Blocks() as demo:
             examples=[
                 [
                     ["./examples/ai_talk.wav", "./examples/bg-image.png"],
-                   "Use the image as the background with a waveform visualization for the audio positioned in center of the video.",
-                   0.7,
-                   0.1,
-                   list(MODELS.keys())[0],
-               ],
-               [
-                   ["./examples/ai_talk.wav", "./examples/bg-image.png"],
-                    "Use the image as the background with a waveform visualization for the audio positioned in center of the video. Make sure the waveform has a max height of 250 pixels.",
+                    "Use the image as the background with a waveform visualization for the audio positioned in center of the video.",
                     0.7,
                     0.1,
-                    list(MODELS.keys())[0],
                 ],
                 [
                     [
+                        "./examples/cat8.jpeg",
                         "./examples/cat1.jpeg",
                         "./examples/cat2.jpeg",
                         "./examples/cat3.jpeg",
                         "./examples/cat4.jpeg",
                         "./examples/cat5.jpeg",
                         "./examples/cat6.jpeg",
+                        "./examples/cat7.jpeg",
                         "./examples/heat-wave.mp3",
                     ],
-                   "Create a 3x2 grid of the cat images with the audio as background music. Make the video duration match the audio duration.",
-                   0.7,
-                   0.1,
-                   list(MODELS.keys())[0],
-               ],
-           ],
-           inputs=[user_files, user_prompt, top_p, temperature, model_choice],
+                    "Generate an MP4 slideshow where each photo appears for 2 seconds, using the provided audio as soundtrack.",
+                    0.7,
+                    0.1,
+                ],
+                [
+                    ["./examples/waterfall-overlay.png", "./examples/waterfall.mp4"],
+                    "Add the overlay to the video.",
+                    0.7,
+                    0.1,
+                ],
+                [
+                    ["./examples/example.mp4"],
+                    "Make this video 10 times faster",
+                    0.7,
+                    0.1,
+                ],
+            ],
+            inputs=[user_files, user_prompt, top_p, temperature],
             outputs=[generated_video, generated_command],
             fn=update,
             run_on_click=True,

requirements.txt

@@ -1,3 +1,3 @@
 openai>=1.55.0
-gradio==5.30.0
+gradio==5.6.0
 moviepy==1