I built a website that makes you hear the voices in your head. They will complete your thoughts, in all the ways they might have unfolded, with your own voice. Everything runs locally in your browser.
I wanted the experience to be 1) local 2) realtime and 3) with voice cloning. Under the hood it’s VAD -> STT -> LLM -> TTS, respectively:
- Silero
- Parakeet CTC 0.6B
- LFM2.5-350
⁃ Pocket TTS
Ofc, they're not technically your thoughts. I finetuned the LLM on Simone Weil’s and Albert Camus’s personal notebooks. It is a small model, dumb in a weird fun way. You can also finetune it on your own notes or messages, and the experience with both your voice and some of your thought-pieces is a bit uncanny. Here’s the repo if you’d like to try: https://github.com/vidalmaxime/voices-in-your-head
I wanted the experience to be 1) local 2) realtime and 3) with voice cloning. Under the hood it’s VAD -> STT -> LLM -> TTS, respectively: - Silero - Parakeet CTC 0.6B - LFM2.5-350 ⁃ Pocket TTS
Ofc, they're not technically your thoughts. I finetuned the LLM on Simone Weil’s and Albert Camus’s personal notebooks. It is a small model, dumb in a weird fun way. You can also finetune it on your own notes or messages, and the experience with both your voice and some of your thought-pieces is a bit uncanny. Here’s the repo if you’d like to try: https://github.com/vidalmaxime/voices-in-your-head