Add results from Hugging Face's Open TTS Leaderboard

#8
by bezzam HF Staff - opened

Congrats on this light-weight model! We've benchmarked it on HF's new Open TTS Leaderboard. I've added some key takeaways that can be mentioned from the model card 🤗

Screenshots can be seen here: https://huggingface.co/datasets/bezzam/tts_leaderboard_screenshots/tree/main/inflect_micro-v2

Hey Eric, thanks again for evaluating Inflect v2 and adding the results!

I’m releasing Inflect v3 soon - it’s a major update over v2 with an expanded 4-model family line (including a larger ~28m parameter TTS model) and I’d really love to get it evaluated on the Open TTS Leaderboard as well. Is there anything I should make sure is included in the repo/model card to make it easy to add? I can send the model over as soon as it’s released.

Thanks!

hi @owensong very cool on the new release! what you've had in the previous model cards has been enough for me to evaluate for the leaderboard.
If you can add a batch inference API, that would be nice! Otherwise I'm looping over each item, and the model's already fast enough 👍 Feel free to ping me when the model is released, so that I can launch the evals!

In 1-2 weeks, we plan to open-source the eval code, which will allow the community to suggest improvements and also new models 🤗

FYI for this model, if you press the "Merge" button below, that will add the TTS leaderboard results to this model card!

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment