| import os |
|
|
| os.environ["PIP_ROOT_USER_ACTION"] = "ignore" |
| os.environ["GRADIO_SSR_MODE"] = "false" |
| os.environ["GRADIO_DISABLE_SSR"] = "1" |
|
|
| import gradio as gr |
| import spaces |
|
|
| RESOURCES = [ |
| { |
| "name": "Vidi (ByteDance)", |
| "category": "Video Model", |
| "description": "ByteDance's multimodal video understanding and generation framework.", |
| "links": [ |
| ("🐙 GitHub", "https://github.com/bytedance/vidi"), |
| ("🚀 Live Demo", "https://vidi.byteintl.com"), |
| ], |
| }, |
| { |
| "name": "Molmo2-8B (AllenAI)", |
| "category": "Open-Weight VLM", |
| "description": "AllenAI's open multimodal vision-language model for rich video and image reasoning.", |
| "links": [ |
| ("🤗 Hugging Face", "https://huggingface.co/allenai/Molmo2-8B"), |
| ("🚀 Playground Demo", "https://playground.allenai.org/?model=molmo2-8b"), |
| ("📝 Blog Post", "https://allenai.org/blog/molmo2"), |
| ("▶️ YouTube Video", "https://youtu.be/Ej3Hb3kRiac"), |
| ], |
| }, |
| { |
| "name": "VideoLLaMA 3 (DAMO-NLP-SG)", |
| "category": "Video-LLM", |
| "description": "Frontier multimodal foundation model tailored for image and video understanding.", |
| "links": [ |
| ("🐙 GitHub", "https://github.com/DAMO-NLP-SG/VideoLLaMA3"), |
| ("🚀 HF Space Demo", "https://huggingface.co/spaces/lixin4ever/VideoLLaMA3"), |
| ], |
| }, |
| { |
| "name": "VideoJev-35B-A3B", |
| "category": "MoE Video-LLM", |
| "description": "35B total parameter (3B active) Mixture-of-Experts model for efficient video understanding.", |
| "links": [ |
| ("🤗 Hugging Face", "https://huggingface.co/ZhuYuhan/VideoJev-35B-A3B"), |
| ("🚀 HF Space Demo", "https://huggingface.co/spaces/hugging-apps/videojev-35b-a3b"), |
| ], |
| }, |
| { |
| "name": "Gemini Agentic Video Understanding", |
| "category": "API & Developer Guides", |
| "description": "Official Google AI Studio and Gemini API guides for agentic video reasoning and frame-level understanding.", |
| "links": [ |
| ("🛠️ AI Studio Developer Guide", "https://aistudio.google.com/learn/agentic-video-understanding-with-gemini"), |
| ("📚 Gemini Video Docs", "https://ai.google.dev/gemini-api/docs/video-understanding"), |
| ], |
| }, |
| ] |
|
|
|
|
| @spaces.GPU |
| def filter_resources(search_query: str, category_filter: str) -> str: |
| query = (search_query or "").strip().lower() |
| sections = [] |
|
|
| for item in RESOURCES: |
| if category_filter != "All" and item["category"] != category_filter: |
| continue |
| if query and ( |
| query not in item["name"].lower() |
| and query not in item["description"].lower() |
| and query not in item["category"].lower() |
| ): |
| continue |
|
|
| links_md = " | ".join( |
| f"**[{label}]({url})**" for label, url in item["links"] |
| ) |
| section = ( |
| f"### {item['name']} `[{item['category']}]`\n" |
| f"{item['description']}\n\n" |
| f"🔗 {links_md}\n\n---" |
| ) |
| sections.append(section) |
|
|
| if not sections: |
| return "### No matching models or resources found." |
|
|
| return "\n\n".join(sections) |
|
|
|
|
| categories = ["All"] + sorted(list({r["category"] for r in RESOURCES})) |
|
|
| with gr.Blocks(title="Video Understanding Models & Demos Hub") as demo: |
| gr.Markdown( |
| """ |
| # 🎬 Video Understanding Models, Demos & Resources |
| Quick-access directory of multimodal video foundation models, Hugging Face checkpoints, interactive playgrounds, and developer guides. |
| """ |
| ) |
|
|
| with gr.Row(): |
| search_box = gr.Textbox( |
| label="Search Models / Links", |
| placeholder="Filter by keyword (e.g., Molmo, VideoLLaMA, Gemini, ByteDance, Vidi)...", |
| scale=3, |
| ) |
| category_dropdown = gr.Dropdown( |
| choices=categories, |
| value="All", |
| label="Category", |
| scale=1, |
| ) |
|
|
| gr.Markdown("---") |
| results_markdown = gr.Markdown(value=filter_resources("", "All")) |
|
|
| with gr.Accordion("📋 Summary Comparison Table", open=True): |
| gr.Markdown( |
| """ |
| | Model / Resource | Category | Model / Repo Link | Live Demo / Space | Extra Links | |
| | :--- | :--- | :--- | :--- | :--- | |
| | **Vidi (ByteDance)** | Video Model | [GitHub](https://github.com/bytedance/vidi) | [Vidi Demo](https://vidi.byteintl.com) | — | |
| | **Molmo2-8B (AllenAI)** | Open-Weight VLM | [Hugging Face](https://huggingface.co/allenai/Molmo2-8B) | [AllenAI Playground](https://playground.allenai.org/?model=molmo2-8b) | [Blog](https://allenai.org/blog/molmo2) · [YouTube](https://youtu.be/Ej3Hb3kRiac) | |
| | **VideoLLaMA 3** | Video-LLM | [GitHub](https://github.com/DAMO-NLP-SG/VideoLLaMA3) | [HF Space](https://huggingface.co/spaces/lixin4ever/VideoLLaMA3) | — | |
| | **VideoJev-35B-A3B** | MoE Video-LLM | [Hugging Face](https://huggingface.co/ZhuYuhan/VideoJev-35B-A3B) | [HF Space](https://huggingface.co/spaces/hugging-apps/videojev-35b-a3b) | — | |
| | **Gemini Video Understanding** | API & Developer Guides | [AI Studio Guide](https://aistudio.google.com/learn/agentic-video-understanding-with-gemini) | — | [API Docs](https://ai.google.dev/gemini-api/docs/video-understanding) | |
| """ |
| ) |
|
|
| search_box.change( |
| fn=filter_resources, |
| inputs=[search_box, category_dropdown], |
| outputs=results_markdown, |
| ) |
| category_dropdown.change( |
| fn=filter_resources, |
| inputs=[search_box, category_dropdown], |
| outputs=results_markdown, |
| ) |
|
|
| if __name__ == "__main__": |
| demo.launch(server_name="0.0.0.0", server_port=7860, ssr_mode=False) |
