#6262nd of 89 in Image and videoFor image, video and design, image and design, video and animation

personalive

Real-time diffusion portrait animation driven by a webcam or video

Stars
4.2k
License
Apache-2.0
Last commit
Aug 2026
Language
Python
Stars, last 29 weeks+1,596
25 Apr 2026: 2,570 stars2 May 2026: 2,594 stars9 May 2026: 3,117 stars16 May 2026: 3,164 stars23 May 2026: 3,195 stars30 May 2026: 3,242 stars6 Jun 2026: 3,277 stars13 Jun 2026: 3,292 stars20 Jun 2026: 3,317 stars27 Jun 2026: 3,346 stars4 Jul 2026: 3,370 stars11 Jul 2026: 3,382 stars18 Jul 2026: 3,390 stars25 Jul 2026: 3,413 stars1 Aug 2026: 3,431 stars8 Aug 2026: 3,458 stars15 Aug 2026: 3,523 stars22 Aug 2026: 3,538 stars29 Aug 2026: 3,579 stars5 Sept 2026: 3,651 stars12 Sept 2026: 3,723 stars19 Sept 2026: 3,783 stars26 Sept 2026: 3,842 stars3 Oct 2026: 3,923 stars10 Oct 2026: 3,966 stars17 Oct 2026: 4,008 stars24 Oct 2026: 4,042 stars31 Oct 2026: 4,066 stars7 Nov 2026: 4,106 stars14 Nov 2026: 4,166 stars

2,570 on 25 Apr 20264,166 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

PersonaLive animates a single reference portrait image from a driving video or webcam, using a streaming diffusion pipeline that can produce arbitrarily long output. It ships a web UI on port 7860, an offline inference script, optional TensorRT acceleration and three-stage training code. It is CVPR 2026 research code, and the README limits use to academic research.

Who it is for: Researchers and tinkerers animating portraits live on an NVIDIA GPU

Strengths

  • Streaming generation produces long videos on 12GB VRAM in offline mode
  • Web UI for live animation; optional TensorRT build gives about 2x speedup
  • Includes three-stage training code and weights on Hugging Face and ModelScope
  • ComfyUI integration exists as a community project

Weaknesses

  • README states academic research use only, limiting practical deployment
  • TensorRT engine should be rebuilt locally (~20 minutes); shipped one is from an H100
  • xformers is incompatible with RTX 50-series; acceleration must be disabled there
  • No Docker files; manual conda, Node.js 18+ and weight download setup

What it needs

  • GPU required
  • Needs conda, Python 3.10, Node.js 18+, xformers or TensorRT (optional)
  • Models: sd-image-variations-diffusers, sd-vae-ft-mse, PersonaLive .pth weights, ONNX, TensorRT
  • port 7860

Also in Image and video

See all 89
Also in Image and video
RankProjectScore
1frigateLocal NVR with realtime object detection for IP cameras37.6k starsMIT93 out of 100
2MoneyPrinterTurboGenerates short videos from a topic, with script, footage, voiceover and subtitles144.3k starsMIT88 out of 100
3toonflow-appInfinite-canvas tool for AI short drama and comic video production22.1k starsMIT88 out of 100
#4InvokeAIWeb UI for diffusion image generation with inpainting canvas and node workflows29.1k starsApache-2.087 out of 100
#5ComfyUINode-graph editor for running diffusion image, video, audio and 3D workflows147.4k starsGPL-3.085 out of 100