https://yummy-fir-7a4.notion.site/dia
Demo with clips. It sounds pretty good, I’m not very familiar with tts so I’m not sure what to expect. 10gb for the full sized model is very reasonable for consumer hardware though. Also really cool this is just two people that made this.
Yeah this is something you can easily run local.
Cool. Though still waiting for a good open source, locally runnable (without needing 64GB of RAM for itself alone) speech to text transcribing model that isn’t awful which I can use to generate subtitles for things from less than perfect audio samples. Doesn’t exist apparently as even Youtube’s transcription isn’t great (though Apple podcast’s transcription is actually really good by comparison).
looks like Dia runs with just 10gb already
The full version of Dia requires around 10GB of VRAM to run. We will be adding a quantized version in the future.
