README.md

metadata

title: Dream Video Face Swap ZeroGPU
emoji: 🎭
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.44.1
python_version: 3.10.13
app_file: app.py
pinned: false
license: other
short_description: ZeroGPU video face and complete head swap
hardware: zero-gpu
models:
  - hacksider/deep-live-cam

Dream Video Face Swap — ZeroGPU

Replace one face in a short video from a single source portrait. The app tracks the selected target identity across frames, preserves the original audio, and exports a browser-friendly MP4.

Features

• Hugging Face ZeroGPU via @spaces.GPU
• Up to 45 seconds per job
• One source portrait → one tracked target identity
• Multiple-person target selection
• 720p, 1080p, and original-resolution modes
• Optional original audio and AI watermark
• Automatic cleanup of temporary outputs
• Public Gradio API endpoint: /swap_video
• GHOST 2.0 complete Head Swap beta (hair and head shape), up to 5 seconds at 720p
• Public Head Swap API endpoint: /swap_head_video

Usage

1. Upload a clear, front-facing source portrait.
2. Upload an MP4 target video.
3. If several people are visible, select the target's left-to-right position in the first clear frame.
4. Confirm that you own the media or have permission from everyone depicted.
5. Click Swap Face.

For complete head replacement, open the Head Swap — GHOST 2.0 Beta tab. Upload a portrait that shows the complete hairstyle and begin with a short, mostly frontal clip. GHOST 2.0 is loaded only when this tab is used, so normal Face Swap startup stays lighter.

For the quickest ZeroGPU run, begin with a 5–10 second 720p clip.

Responsible use and licensing

Do not use this app for impersonation, fraud, harassment, deceptive political content, or non-consensual sexual content. You are responsible for obtaining the necessary permissions.

The application code is provided for research and demonstration. Face-engine and GHOST 2.0 weights are subject to their upstream terms. Review those licenses before commercial use.