Spaces:
Running
Would a PR adding "long-form" sentences be accepted?
Hey there, I'm interested in comparing TTS models for long-form generation (ie 1-5 minutes of text). Listening to 10+ minutes is probably pushing the internet's attention span, but I think 30-second snippets of text (so listening to 1 minutes worth of audio to cast a vote) would work. I'd personally be happy to cast a few hundred votes, but at that point I'm probably skewing the dataset (:
If I compiled a dataset of text snippets, lets say ~100 words each or about 30s of audio, would that PR be accepted? From what I can see, I'd just need to add the file and then ammend the first few lines of app.py to read in the text file and append it to sents.
Thanks for the arena!
Sure, I'd accept it, if an extra checkbox has to be used for the random sentence picker to choose them.
This month I will merge the structured branch which will increase the Arena code's readability and closer match the original Arena.
https://huggingface.co/spaces/Pendrokar/TTS-Spaces-Arena/tree/structured
Code restructured, this should make things easier for those that duplicate this Space