Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Hi, I just want you to know that I'm blown away by your keyboard. I'm not even a native English speaker and it's incredible. After paying around with it for ten minutes I decided to pay the 5 dollars to show my support to this initiative.

I'm especially happy to see that you mdocumented how to use your own finetuned models and predictive text transformer. I'm betting it will end up enabling community driven multi-lingual dictionaries and transformers. Thanks a lot!

One feature request I have though is (although I understand the limitations of transformers) to seamlessly handle more than 30s of audio, even though it takes more time to transcribe it still is a huge pain to have to wait for the first to end writing to then speak again. It's not that difficult to queue audio segments right :)?

Also, the language choice picker from the keyboard shows the choices too high : on tall phones it's cumbersome to have to reach up there.

Also the beamsearch is "recommended" but it would be better to also mention that it makes the decoding way faster.

Overall, allowing more access to advanced settings would be awesome for tinkerers, for example choosing the depth of the beam search!

Anyway thanks a lot FUTO is awesome!



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: