m2024's comments

On: The lumbar puncture that changed my life

[m2024]

My deepest condolences. I will have a lumbar puncture done in a few hours and am facing a diagnosis that could be extremely grave, but everywhere I look there are stories of people who have had it harder. I will keep you and your wife in mind today, and I hope you find peace.

[rpaddock → m2024]

Thank you.

Make sure the puncture location is well documented. That is a large part of the battel.

On: China is quietly reducing its reliance on foreign ...

[m2024]

No one who has been to China would ever think that. It's an awe-inspiring country with massive cities and fascinating people.

On: China is quietly reducing its reliance on foreign ...

[m2024]

To "steal" this sort of information and utilize it requires immense intelligence.

On: BASE TTS: The largest text-to-speech model to-date

[m2024]

Check out `whisper` and `whisper-cpp` for ASR.

I am running the smaller models in near real-time on a 3rd gen i7, with good results even using my terrible built-in laptop mic from a distance. The medium and large models are impressively accurate for technical language.

[qwertox → m2024]

I'm using Whisper to transcribe notes I record with a lavalier mic during my bike rides (wind is no problem), but am using OpenAI's service. When it was released I tested it on a Ryzen 5950x and it was too slow and memory hungry for my taste. Using large was necessary for that use case (also, I'm recording in German).

[kkielhofner → qwertox]

The original release was full precision model weights running in an old version of PyTorch with no optimizations.

Fast forward to now and you have faster-whisper (using Ctranslate2) and distil-whisper optimized weights.

Between the two of them Whisper Large uses something like 1/8th the memory and is likely at least an order of magnitude faster on your hardware.

German has no effect on these metrics and for accuracy it actually has a lower word error rate than English.

[GaggiX → qwertox]

With Whisper, you can find many smaller models that are fine-tuned for a particular language, so even smaller models can perform adequately.

[jamil7 → m2024]

Whisper is for STT though right?

[qwertox → jamil7]

The term STT is not used, it's called ASR, Automatic Speech Recognition. I mean, I was referring to both TTS and ASR in my comment.

[ametrau → qwertox]

Not used by who? It’s a better term. Let’s use it.

[m2024 → jamil7]

I also use STT but the parent poster wrote ASR so for clarity I responded in kind.