Hacker News Reader

Comment by m2024 | Hacker News Reader

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.

View on HN

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).

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

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.

Whisper is for STT though right?

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.

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

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

Reply on news.ycombinator.com