Azerbaijani Whisper Turbo GGML for whisper.cpp
This repository provides a
ggml model file for
whisper.cpp, converted from
the original
LocalDoc/azerbaijani-whisper-turbo
Transformers checkpoint.
What this is
- Source model:
LocalDoc/azerbaijani-whisper-turbo
- Converted format:
GGML .bin
- Target runtime:
whisper.cpp and wrappers built on top of it
- Intended language: Azerbaijani speech recognition
This repo exists because the original LocalDoc model is published in the
Transformers ecosystem, while some mobile and embedded runtimes expect a
whisper.cpp-compatible ggml model file instead.
Conversion note
The file in this repository was converted from the original LocalDoc checkpoint
for use with whisper.cpp. In practical terms:
- the original model was downloaded from Hugging Face
- tokenizer files required by
whisper.cpp were exported
- the checkpoint was converted into a
ggml binary
The conversion was prepared for on-device inference experiments in Flutter using
whisper_ggml_plus, which itself uses whisper.cpp under the hood.
Intended use
Use this model if you need the Azerbaijani Whisper Turbo checkpoint in a
whisper.cpp-compatible format.
Typical use cases include:
- local speech-to-text on desktop or mobile
- Flutter applications using
whisper_ggml_plus
whisper.cpp command-line or embedded integrations
File
ggml-azerbaijani-whisper-turbo.bin
Original model
Please refer to the original source model for upstream training details,
evaluation context, and authorship:
Notes
The Turbo variant is substantially larger than the Small variant. It may be
more suitable for optional downloads or test builds than for bundling directly
into mobile application assets.
Acknowledgements
- LocalDoc for the original Azerbaijani Whisper model
- OpenAI for Whisper
whisper.cpp contributors for the C/C++ inference runtime
