hitz-zentroa/whisper-lm-transformers

Add n-gram and LLM language model support to HF Transformers Whisper models.

42
/ 100
Emerging

This project helps improve the accuracy of speech-to-text transcription, especially for languages with fewer data resources. It takes an audio file and a pre-trained Whisper model, along with a custom language model (either n-gram or a large language model), and outputs more accurate text transcripts. Researchers, linguists, and engineers working with automatic speech recognition, particularly for less common languages, would find this tool useful.

No commits in the last 6 months. Available on PyPI.

Use this if you need to enhance the transcription quality of an existing Whisper speech-to-text model, especially for specific languages or domains where standard models might struggle.

Not ideal if you simply need basic speech-to-text transcription without needing to fine-tune or improve accuracy with custom language models.

speech-to-text language-technology linguistics audio-transcription low-resource-languages
Stale 6m
Maintenance 2 / 25
Adoption 5 / 25
Maturity 25 / 25
Community 10 / 25

How are scores calculated?

Stars

14

Forks

2

Language

Python

License

Apache-2.0

Last pushed

May 06, 2025

Commits (30d)

0

Dependencies

8

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/hitz-zentroa/whisper-lm-transformers"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.