SoulX-FlashTalk
by Soul-AILab
14B open-weights model for real-time, audio-driven talking-avatar video (image to video).
AI presenters and avatars Open source
1. Price: huggingface.co
Open weights, free to download; self-hosted.
huggingface.coNo quote yet: price is read from the official pages.
2. Commercial use: huggingface.co
“license: apache-2.0”
Apache 2.0 allows commercial use of the code and weights; keep the license and notices, and you are responsible for rights to the faces and voices you use.
huggingface.co3. Credit the author: huggingface.co
huggingface.co4. Agent access: github.com/Soul-AILab/SoulX-FlashTalk
There is no package or API. Files are downloaded by hand from the official site.
github.com/Soul-AILab/SoulX-FlashTalkLicense checked Oct 27, 2026 against 1 official page. Tap a verdict to see its source. How we checked Report an error
SoulX-FlashTalk: a talking video from one photo and audio
SoulX-FlashTalk is a free, open AI model that turns one portrait picture and a voice recording into a video of that face talking in sync with the audio. It is built for live, endless streams, but it is very large and needs data-center graphics cards. The model and code are under the Apache 2.0 license, which allows paid use.
Where you can use it
A live talking avatar
Put a speaking face on a voice assistant or chat, so the picture moves its lips while the reply plays, with little delay.
Live streams and video podcasts
Run a virtual host for a stream or a video podcast that talks for as long as the audio goes on.
Long videos without cuts
Make a long talking-head video, such as a lecture or a narrated story, where the face stays the same from start to end.
Cartoon and animal characters
Make a drawn character, a painted portrait or an animal picture sing or speak, for a fun clip or a mascot.
Presenter for course videos Idea
Pair a recorded or generated voice-over with a presenter photo you have rights to, for lessons or product explainers.
Research on digital humans Idea
Study or compare talking-avatar methods with an open model and its technical report, instead of a closed paid service.
As reported by Soul-AILab: 32 frames per second and a 0.87-second start-up delay for the 14B model; Training converged in 1,000 steps of fine-tuning and 200 steps of distillation, a 23x efficiency gain.
SoulX-FlashTalk popularity
How to start
Tell your AI assistant
“On a Linux machine with an NVIDIA GPU of at least 40 GB, set up SoulX-FlashTalk from github.com/Soul-AILab/SoulX-FlashTalk as in its README, then run generate_video.py with --cpu_offload, my portrait as --cond_image, my 16 kHz WAV as --audio_path and --save_file out/talk; use only a face I have permission to animate.”
Show codeStep 1 or Step 2 or Step 3
Make a Python 3.10 environment and install PyTorch 2.7.1 for CUDA 12.8, then the other requirements, flash-attention and FFmpeg.
conda create -n flashtalk python=3.10
conda activate flashtalk
pip install torch==2.7.1 torchvision==0.22.1 --index-url https://download.pytorch.org/whl/cu128
pip install -r requirements.txt
Download the 14B model and the audio encoder it needs from Hugging Face.
pip install "huggingface_hub[cli]"
huggingface-cli download Soul-AILab/SoulX-FlashTalk-14B --local-dir ./models/SoulX-FlashTalk-14B
huggingface-cli download TencentGameMate/chinese-wav2vec2-base --local-dir ./models/chinese-wav2vec2-base
Run the single-GPU script; it needs more than 64 GB of graphics memory, or about 40 GB with the --cpu_offload switch.
bash inference_script_single_gpu.sh
Mistakes to avoid
The model needs more than 64 GB of VRAM on one card, or 40 GB with offload, and the install pins CUDA 12.8 wheels.
The audio is turned into features by this separate wav2vec model before the video is made.
The script hard-codes the example files examples/man.png and examples/cantonese_16k.wav.
The README limits real-time speed to large multi-GPU setups.
The README lists FFmpeg as a separate install step; the video with sound is written through it.
License
Apache-2.0 license. Apache 2.0 for the code and the model weights, so paid use is allowed. Keep the license and notices when you share the files, and say which files you changed.
“license: apache-2.0” huggingface.co
“license: apache-2.0” huggingface.co
You may share the weights or code if you include the license, keep notices and mark changed files.
“You must give any other recipients of the Work or Derivative Works a copy of this License” LICENSE on GitHub
AllowedWith conditionsNot allowedNot stated or not checked yet
Is SoulX-FlashTalk free for commercial use? The full answer Not legal advice. Check the license page before you ship.
For your agent
License rules tell your agent what it may do with SoulX-FlashTalk. Paste them into your prompt or AGENTS.md.
- SoulX-FlashTalk (AI presenters and avatars). 14B open-weights model for real-time, audio-driven talking-avatar video (image to video). Price: free; commercial use OK.
Connect: download from https://huggingface.co/Soul-AILab/SoulX-FlashTalk-14B (no API).
Rule: Keep the Apache-2.0 license and notices when you redistribute the weights.
Rule: Do not clone a real person without their consent; the model card gives no output license, so check the repo.
License: https://huggingface.co/Soul-AILab/SoulX-FlashTalk-14B/blob/main/README.md (checked 2026-10-27; details: https://agentgid.com/assets/soulx-flashtalk/)
Alternatives
SoulX-FlashTalk This page
Pick LivePortrait to make a photo copy the face moves of a driving video rather than follow audio; it runs on smaller machines.
Pick HeyGen for ready avatar videos from a script in an app or an API, paid per minute, with no graphics cards of your own.
Pick Tavus for live talking-video conversations as a hosted API instead of running a 14B model yourself.
VOCA (Voice Operated Character Animation)
Speech-driven 3D facial animation model with code; data and trained model need a free sign-in.
AI avatar video platform with an API and MCP; free videos carry a watermark.
Questions
Can I run it on my laptop or a gaming PC?
No. One card needs more than 64 GB of graphics memory, or about 40 GB with --cpu_offload; the team points to SoulX-FlashHead for consumer cards.
What do I give it?
A portrait image, a speech audio file and a short text prompt; the example uses a PNG and a 16 kHz WAV.
Where does the video go?
To the file set with --save_file, or into a sample_results folder by default.
Can I use it commercially?
The model and code are under Apache 2.0, which allows commercial use; keep the license and notices with any copies.
What is it built on?
On the InfiniteTalk and Wan models, with code from Self Forcing.