Any way to extract punctuation marks and transcription

[gormat]

Hey.
Really cool project !!!
I am using silero-stt model to extract text.
I just have few questions:

• This model does not extract punctuation marks and does not split into sentences, am I missing something?
• Is there a way to get start and end timestamps for each word as well?

Thank you.

[snakers4]

Hi @gormat,

This model does not extract punctuation marks and does not split into sentences, am I missing something?

in most of cases written punctuation is not related to how people actually speak (except for maybe ?)
in the CE version of our models we just provide the STT part
but typically the pipeline may work as follows:

• you can use some VAD to split speech into utterances. typically people separate ideas with pauses
• you can tune some pre-trained LM like a transformer to add capital letters, commas, full-stops etc

Is there a way to get start and end timestamps for each word as well?

• Please use this colab
• Proceed to PyTorch example
• Search for comment # align example

[snakers4]

you can tune some pre-trained LM like a transformer to add capital letters, commas, full-stops etc

We actually published models for this