Spaces:
Running
on
CPU Upgrade
How open source architectures compares
Hello,
I would be very much interested in testing various other non-commercial TTS for researchers, like VITS, Matcha-TTS, Grad-TTS, Glow-TTS, Tacotron 2, FastSpeech etc. Is there any plans to evaluate those or anyway one can help with adding these?
Hi @shivammehta25 - Lovely to see you here! we'll gradually add models based on community votes (we'll open it tomorrow). We don't want to flood the arena with 10+ models in one go as each model needs to receive up to 700 unique votes to show up on the leaderboard. In general, we're biased towards recent + open access models which have been trained on more than just LJSpeech or VCTK.
Do you have any specific recommendations for models that fit the above bill? :)
Also, we have limited compute for hosting models, we're looking for ways to better serve more models.
Cheers!
Yeah, that makes sense, I wonder if TorToiSe or Bark are in the pipeline to be tested?
Yes! In pipeline for sure! 🤗
@shivammehta25
I created the #30 pull request and you can clone the Space to add the TTS you wanted. That is if the TTS has a 🤗 Space, just add it, with parameter overrides if needed, right around here:
https://huggingface.co/spaces/Pendrokar/TTS-Spaces-Arena/blob/main/app.py#L54-L65
(closing this since it has been discussed and in the pipeline)