📱 NanoAvatar

Changing how often avatar applications can be used and how many users they can serve

📱 Snapdragon 8 Gen 3 · Lite 37 FPS · ⚡ 112 ms model first frame

📱 Android downloads (Full / Lite) · 💻 Code and usage

🎬 Demos · 📊 Benchmarks · English · 简体中文

Try it offline, without an API key. The app includes the models and an avatar. Record your voice to animate it. Android guide.

🎬 Demos

Generalization demo: The examples below were not used to train the model, nor was it specifically fine-tuned for them.

English demo	Chinese demo

Generate high-fidelity talking-avatar videos directly on your phone, without a cloud GPU.

📊 Performance

Model	Device	Best model FPS	Best first frame	Memory / VRAM estimate
📱 NanoAvatar	Z60 Ultra · Snapdragon 8 Gen 3 Android 14	35 FPS	127 ms	834 MiB
📱 NanoAvatar Lite	Z60 Ultra · Snapdragon 8 Gen 3 Android 14	37 FPS	112 ms	700 MiB
⚡ Quantized	RTX 4090 Windows CUDA	333 FPS	18 ms	834 MiB
🖥️ Full precision	RTX 4090 Windows CUDA	224 FPS	37 ms	1119 MiB

Android first-frame timing runs from audio feature extraction to the first generated image.

The model generates a 256 × 256 face region, composited at the source video's original resolution.

🌊 Streaming generation: start speaking as audio arrives. With a streaming LLM and streaming TTS, NanoAvatar starts speaking in about 0.3 seconds in our tests, without waiting for the complete audio.

📦 Choose and download a model

Model	Files	Contents
⚡ Quantized	quantized/	PyTorch/CUDA: HuBERT W8A16, mixed INT8 lip-sync model and CUDA DLL
🖥️ Full precision	full-precision/	PyTorch: HuBERT FP16 and lip-sync model FP32
📱 Android compiled	android-qnn/	CPU: original FP32 face/audio encoders. QNN: HuBERT W8A16 and HTP FP16 generator

The quantized CUDA runtime targets Windows CUDA SM89 (RTX 4090). The Android package targets Snapdragon 8 Gen 3 / HTP v75.

Install the Hugging Face CLI, then download the package you need:

python -m pip install huggingface_hub

⚡ Quantized

hf download wpydcr/NanoAvatar --include "quantized/*" --local-dir models

🖥️ Full precision

hf download wpydcr/NanoAvatar --include "full-precision/*" --local-dir models

📱 Android compiled

hf download wpydcr/NanoAvatar --include "android-qnn/*" --local-dir models

Keep the complete directory structure when downloading.

Running the models and integrating them: NanoAvatar on GitHub.

License

chinese-hubert-large and its quantized and compiled versions retain the upstream MIT license.

Our lip-sync model and its quantized and compiled versions use CC BY-NC 4.0. Academic research and other noncommercial uses are permitted, subject to attribution and the other license terms.

Commercial licensing: wupingyu@mail.ustc.edu.cn.

See LICENSE for the full terms.