Expression transfer
animate still photo with your live face video
// official site: github.com ↗
AVATAR / VIDEO · PRO TIER
LivePortraitpro
LivePortrait is Kuaishou's real-time portrait animation model — drive a still image with another person's face video, transferring expression and head motion frame-by-frame. Higher quality and faster than SadTalker for the same use case.
// What it is
A closer look.
LivePortrait is Kuaishou's real-time portrait animation model — drive a still image with another person's face video, transferring expression and head motion frame-by-frame. Higher quality and faster than SadTalker for the same use case.
The 2024 evolution of synthetic talking-head video, optimized for real-time deployment.
// Use cases
What it's for.
Concrete scenarios where teams pick LivePortrait over the SaaS alternative.
Eye / lip retargeting
fine-grained control modes
Real-time animation
30 FPS on RTX 4090
Cross-identity animation
your face acting → applied to another portrait
Game character animation
from concept art + voice acting
Avatar generation
for personal video content
// Who it's for
Built for these teams.
If your team profile matches one of these, LivePortrait is a strong fit out of the box.
Animation studios
retargeting facial performances
Game studios
lipsync from voice + driving video
Educational creators
with branded character avatars
Personal video creators
producing animated content
VR / metaverse builders
for avatar systems
// Differentiators
Why teams pick LivePortrait.
When evaluating self-hosted options for this category, here are the dimensions on which LivePortrait consistently lands above the alternatives.
- ✓MIT license (code) — fully open
- ✓Higher quality — than SadTalker at similar VRAM
- ✓Real-time speed — (30 FPS on RTX 4090)
- ✓Multiple retargeting modes — eyes only, eyes + lips, full expression
- ✓Kuaishou backing — major Chinese video platform's research arm
- ✓Active development — + community LoRAs
// Integrations
Connects to.
The stack you'll plug LivePortrait into — services, protocols, and adjacent apps in the BluixApps catalog.
◇
Gradio web UI
included
◈
Pair with
XTTS / F5-TTS for audio → driving video → portrait (advanced workflow)
◆
Pair with
SDXL / Flux for source portrait generation
▣
ComfyUI nodes
community wrappers
▦
CLI mode
for batch
// Adoption & deployment
Notable users & community
- 14k+ GitHub stars
- Kuaishou (KwaiVGI) research backing
- Featured in mid-2024 synthetic video roundups
- Active research community + commercial early adopters
- Multiple ethical-deployment workflows published
What we ship
- Cloned
KwaiVGI/LivePortraitrepo - pytorch CUDA 12.4 base + ffmpeg
- Gradio app launcher
- Persistent volumes: repo, pretrained_weights (~2 GB), output (MP4)
- Port 7875 mapped (Gradio default 8890)
- Install report at
/root/bluixapps/liveportrait.txt - Acceptable Use Policy prominently noted
- LivePortrait vs SadTalker positioning explained
- Pairing suggestions (XTTS for audio, SDXL for portrait, advanced workflow notes)
- GPU pre-flight check via
bluixapps_ensure_nvidia_runtime - Backup hook covers weights + outputs
// Tips & operations
Run it properly.
Operational guidance from running this in production — what to lock down, what surprises people.
// PERFORMANCE
Source portrait
high-res, frontal, single face
// SECURITY
Driving video
your face acting (or licensed footage), clear lighting
// OPERATIONS
Modes
// RELIABILITY
Resolution
256×256 default; 512×512 supported (slower)
// DEPLOYMENT
Speed
30 FPS on RTX 4090, ~5-15 FPS on RTX 3060
// SCALING
VRAM
8 GB GPU recommended
12288
// min ram (MB)
12
// min disk (GB)
7875
// access port
http
// protocol
pro
// bluixapps tier
Project resources
Official sitegithub.com ↗