rcheung/longcat-avatar-ca_en
1
0
Fork 0
mirror of https://github.com/rebots-online/longcat-avatar-ca_en.git synced 2026-05-20 20:31:41 -04:00
Code Issues Projects Releases Packages Wiki Activity
from huggingface gradio
2 commits 1 branch 0 tags 38 KiB
Find a file
Exact
Robin L. M. Cheung, MBA 8b259563bd Add thorough README, architecture docs, and mirror config. Defer llmfit/lmagent integration.
2026-03-22 09:45:44 -04:00
DOCS Add thorough README, architecture docs, and mirror config. Defer llmfit/lmagent integration. 2026-03-22 09:45:44 -04:00
.env Add thorough README, architecture docs, and mirror config. Defer llmfit/lmagent integration. 2026-03-22 09:45:44 -04:00
app.py Initial commit – English‑only Gradio demo for LongCat‑Video‑Avatar 2026-03-22 09:29:13 -04:00
CHECKLIST.md Initial commit – English‑only Gradio demo for LongCat‑Video‑Avatar 2026-03-22 09:29:13 -04:00
README.md Add thorough README, architecture docs, and mirror config. Defer llmfit/lmagent integration. 2026-03-22 09:45:44 -04:00

README.md

LongCat Avatar Gradio (English)

This repository hosts a self-contained, single-file Gradio demo for the LongCat-Video-Avatar model. It is designed to run locally or on HuggingFace Spaces (ZeroGPU environment).

Features

  • Automatic Dependencies: Installs only necessary packages at runtime.
  • Auto-Download: Fetches the LongCat-Video source code (via ZIP) and model weights directly from the HF Hub.
  • Config Patching: Automatically patches the model configuration to replace flash-attn (FlashAttention) with sdpa for better compatibility in environments where flash-attn cannot be easily installed.
  • Audio-to-Video Generation: Supports both Audio-Text-to-Video (AT2V) and Audio-Image-to-Video (AI2V).

Quick Start

You do not need to clone the original LongCat-Video repository. Just run the app.py script:

# 1. Install minimum requirements (or let the script install them)
pip install gradio requests huggingface_hub spaces

# 2. Run the Gradio UI locally
python app.py

Open the local URL provided by Gradio (usually http://127.0.0.1:7860) in your browser.

Repository Mirrors

  • GitHub (Primary): https://github.com/rebots-online/longcat-avatar-ca_en.git
  • Forgejo (Mirror): https://git.robin.mba/rcheung/longcat-avatar-ca_en

The Forgejo repository is configured as a pull mirror. Any pushes to the GitHub repository will automatically be synchronized to Forgejo.

Forgejo API Token (Optional)

If you need to perform manual Git operations directly against the Forgejo remote, you can export your API token. A .env file is provided in this repository for convenience:

export FORGEJO_API_TOKEN=5cb8182c2596e5e83889c116b1cb563e6da06a37
source .env

(Note: Pushing to GitHub is the recommended workflow to ensure both remotes stay in sync).

Deferred Integrations

Integration of the llmfit and lmagent memory footprint size optimizers has been deferred pending access to the respective packages. Once provided, pre-flight memory optimization hooks will be integrated into the model loading pipeline.