Github :

****https://github.com/dataX-ai/voice-studio-frontend-2vid

https://github.com/dataX-ai/voice-studio-backend-2vid

Link to Binaries

Link to dmg file: https://github.com/dataX-ai/voice-studio-frontend-2vid/releases/latest/download/VoiceStudio.dmg - We will update the signed dmg soon. It’s in the process of getting notorized by Apple.

In our product 2Vid there is a huge demand for better voices of the AI UGC’s. We mainly integrate with opensource models. Every week there seems to be a new breakthrough in the space and a better model is released.

https://youtu.be/nVI_eAZEAlM

We have struggled to playaround with opensource voice models be it for TTS, STT or voice cloning. For SLMs(Small Language Models) there are tools like Ollama, LM Studio, that allows users to inference these voice models locally on their consumer hardware.

So we thought that we should build something to do inference and prototyping of voice models easily on consumer hardware. Hence the idea of Voice Studio was born inspired by LM Studio

We are building this open-source as a electron based desktop app for Mac, linux and windows.

The backend is in python that runs locally on the consumer hardware via docker.

Future Work

• Currently only kokoro is active as a model in the product. We need to load more, Next up is integrating llasa tts.
• Do awareness drive on the product and get it to the hands of users to see if they really find use in this.
• Build separate runtimes for separate systems like for Mac we can build a MLX based runtime to improve inference speed and reduce memory footprint.