OpenAI’s Whisper is AMAZING!

[@Param3021]

Model is looking great! Thank you for sharing :D

[@underfitted]

Sure thing!

Thank you for the instructions and the colab notebook. Surprisingly, it works for Turkish as well.

[@underfitted]

Awesome!

[@cygmoid]

That's really amazing. Nice work Santiago

[@underfitted]

Thanks!

[@PritishMishra]

The quality of your videos is GREAT! your channel will blow up soon! Edit: May I ask which font you have used at 5:10?

[@underfitted]

Thank you very much! That’s my handwriting :) I did it on an iPad.

[@joaodiogocosta]

Please keep sharing!

[@underfitted]

Will do!

[@dakshbhatnagar]

Daaayuuumm, that was quick and easy!

[@underfitted]

Yessir

[@ThomasConover]

I have tried this AI now a few hours and its incredibly impressive so far. It is super slow without a GPU tho.

[@underfitted]

Yeah to both of those comments. You need a GPU is you want to do something useful with it.

[@MyHowHowHow]

You can get great performance on CPU if you use the fork called WhisperCpp.

[@jayperalta7104]

this video is AWESOME. I do have a question though, How and where do I upload my prerecorded audio file? Thanks,

[@underfitted]

You can upload it to Colab. There's a file view on the sidebar. You can drag and drop files there to upload them. Keep in mind they will be deleted when the Runtime is restarted.

[@canozturk369]

great work, great video. thanks

[@underfitted]

Thanks!

[@SergioEspallargas]

Love it!

[@underfitted]

Thanks :)

[@ricosrealm]

Hey Santiago, this video and your Diffusion video have a stereo audio imbalance between left and right. Realized it when listening with headphones.

[@underfitted]

Yeah. Pretty crappy audio. I think I fixed it in my last few videos.

[@jamges]

Can I use the whisper api to transcribe sound errors? I’m a speech therapist so I want to see where the user makes an error “wabbit” instead of “rabbit” without whisper autocorrecting

[@ashwinshetgaonkar6329]

thanks for sharing

[@underfitted]

Absolutely!

[@sigib911]

Amazing, so how can I used in realtime? that would be even great

[@avnishmishra2770]

Man, thanks for this Litt video❤

[@underfitted]

Definitely!

[@abandonedmuse]

OMG ARE YOU CUBAN TOO? I live in miami too! Yeah! Oye los Cubanos estamos en todo!

[@underfitted]

Ha ha! Verdad que si!

[@wimdenherder]

Do you use this for your own subtitles too?

[@underfitted]

I don’t. Yet

[@aleksandrszagorskis3306]

Playback speed 1.5

[@underfitted]

Well, that will certainly go by quick! :)

[@mgreek31]

Good 👍

[@RobinGlauser]

Is there a way to fine tune it? It would be awesome to be able to fine tune the german part to also understand swiss german :) Chuchichästli

[@underfitted]

Good question! I think there is, especially because they open sourced the model.

[@amphem]

do French know about it?

[@underfitted]

It does know French, yes.

[@nickkatsaras9721]

Nice video, but putting your hand on the camera is a bad habit.. maybe consider not doing that anymore, it's really annoying. Otherwise, great.

[@underfitted]

Do you mean covering the lens during scene transitions? I never heard this was annoying before. Do you mind elaborating a bit? And thanks very much for the feedback! It helps a ton to make better videos!