Ask HN: Best speech-to-text resource?

22 points by jamesandthewolf 4y ago ↗ HN

Anyone know any good speech to text resources Ive tried a few but they keep not writing down the correct words looking more free and open source links.

I have terrible spelling and grammar as I suffer from Dyslexia and the speech to text kind of helps me. I can write but it's harder for myself then speaking. I can read ok my the spelling part is difficult as I mix words and letters up and spend most of my time just checking and rewriting large portions of my narratives.

9 comments

[ 3.0 ms ] story [ 38.8 ms ] thread

geenat 4y ago ↗

If it's really important to you, latest pixel phone honestly. Google has an excellent implementation here.

GhettoComputers 4y ago ↗

This app probably uses the same APIs and it’s free. Not as integrated into listening into the apps though.

https://play.google.com/store/apps/details?id=com.google.aud...

innerzeal 4y ago ↗

Otter.ai is pretty good and free mins are available.

Cilvic 4y ago ↗

https://github.com/coqui-ai/STT

woodson 4y ago ↗

If you're fine writing code to build a solution yourself, try the Conformer or ContextNet models in Nvidia NeMo (https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en...) or Vosk (https://alphacephei.com/vosk/), which nicely packages an API for Kaldi chain/LF-MMI models.

GhettoComputers 4y ago ↗

Is it only English? I used this and it’s amazing. Google live transcribe is great!

https://play.google.com/store/apps/details?id=com.google.aud...

jamesandthewolf 4y ago ↗

Thanks I'll check these out

runnerup 4y ago ↗

https://www.rev.ai/ beats Google's speech-to-text models quite impressively.

jlalfonso21 4y ago ↗

If you can code or know how to use some demos, you can give a try to Vosk, this is a opensource project with multiple implementations and language models, opensource as well, all of this offline. They have smalls and big models, for mobile apps, iot, asterisk, and much more

https://alphacephei.com/vosk/