LongCat has officially launched a new open-source talking-avatar model, widely regarded as achieving state-of-the-art (SOTA) performance. Released under the permissive MIT license, the model allows the developer community to widely customize and integrate it into commercial products.
Background
This release marks a significant milestone in the field of facial video generation. The development team has also hosted a Hugging Face Space, enabling the community to try out the demo directly. With the MIT license, legal barriers are stripped away, paving the way for innovative projects spanning from education to entertainment.
Why This Matters
The model's impressive capabilities open up various new business opportunities, such as building life-like AI tutors, automated video dubbing systems, or interactive coding agents capable of video communication. Having such a high-quality model open-sourced makes it much easier for Vietnamese startups to access and customize, freeing them from dependence on expensive proprietary APIs.