common-voice VS NeMo

Mozillas Common Voice Project (https://commonvoice.mozilla.org/) is creating an open dataset for many minority languages to make it easier to support them in STT systems. If you speak one of these languages please consider donating a few minutes of your voice.

> it was not at all obvious to me there was some way of speeding up getting a language in the first place.

Yeah, that's the biggest failing of Common Voice in my opinion. Getting a new language up to speed could be much improved by simply adding a few links to documentation, but even the existing links are broken, which I reported in March 2022... https://github.com/common-voice/common-voice/issues/3637

> I have no interest in wasting time contributing to a UI translation I actively don't want to be subjected to

Translating the UI may still help you get other people to record, even if you don't want to use it yourself.

> I'll see if I can submit some sentences at least

If you want to go faster, there's also a project to extract sentences from Wikipedia etc. in small doses Mozilla's lawyers and Wikimedia's lawyers have agreed are fair use. I think you'd only need to define how Norwegian Bokmål separates sentences. (E.g. after a period but not if it's a common abbreviation like "etc." in the preceding sentence.)

There is a lot of TTS and SST development going on (https://github.com/mozilla/TTS; https://github.com/mozilla/DeepSpeech; https://github.com/common-voice/common-voice). That is the only way they work: Contributions from the wild.

Since you're interested in STT and TTS, let me just plug in Mozilla's Common Voice, a way for everyone to contribute to an open source data set for STT. You can record yourself or verify other people's recordings!

[커먼보이스 전자우편](mailto:[email protected]) || Common Voice || Korean Language Homepage || FAQs || Speaking Aloud and Reviewing Recordings || Sentence Collector || NVidia/NeMo

How can it be used for transcription?

In their website I only see an interface for either uploading audio or submitting transcriptions:

https://commonvoice.mozilla.org/es

The Github repo they mention (https://github.com/common-voice/common-voice) seems to be just that sample collection software. I do not see where I can download the software to transcribe audio.

A practical example: in the speech community, getting good results on datasets such as TIMITwould not be revolutionary, since it's super small and very old. On the contrary, things like Common Voice (which is constantly crowd-sourced) would be much more impactful. Just my two cents :)

I don't test WaveRNN but from the ones that I know the best that is open source is FastPitch. And it's easy to use, here is the tutorial for voice cloning.

Nowadays... bigger repos like https://github.com/NVIDIA/NeMo are all pytorch, lots of work also published by Meta and Microsoft is all torch. I check new work on GitHub all the time and I haven't seen a Tensorflow repo in years except one.

- https://github.com/NVIDIA/NeMo

Can’t you use the NeMo model and just strip the punctuation from the output again if you don’t want it? You can also fine tune the the model with capitalization only if you look at the examples https://github.com/NVIDIA/NeMo/blob/stable/tutorials/nlp/Punctuation_and_Capitalization.ipynb The capitalization and punctuation are annotated separately (U indicates that the word should be upper cased, and O - no capitalization ). The model seems to be a token level classifier not seq to seq so there should also be a way to get just the capitalization part but you would have to look into the model as it’s not shown in the examples.

I think there's been talk to do speaker diarization with whisper-asr-webservice[0] which is also written in python and should be able to make use of goodies such as pyannote-audio, py-webrtcvad, etc.

Whisper is great but at the point we get to kludging various things together it starts to make more sense to use something like Nvidia NeMo[1] which was built with all of this in mind and more

[0] - https://github.com/ahmetoner/whisper-asr-webservice

[1] - https://github.com/NVIDIA/NeMo

[커먼보이스 전자우편](mailto:[email protected]) || Common Voice || Korean Language Homepage || FAQs || Speaking Aloud and Reviewing Recordings || Sentence Collector || NVidia/NeMo

The final security grain was added with speech-to-text anti-spoofing built on QuartzNet from the Nemo framework. This model provides a decent quality user experience and is suitable for real-time scenarios. To measure how close what the person says to what the system expects, requires calculation of the Levenshtein distance between them.