AI GitHub
faster-whisper for CTranslate2 transcription
Run Whisper recognition through CTranslate2 with configurable compute precision.
Key Takeaways
- The implementation uses converted Whisper checkpoints and exposes segment-level transcription in Python.
- Install a CTranslate2 configuration compatible with your accelerator and chosen precision.
- Run a folder of representative clips with consistent decoding settings.
- Libraries and accelerator requirements depend on the compute path.
How it works
The implementation uses converted Whisper checkpoints and exposes segment-level transcription in Python. Precision and batching options let the same workload be tested under different resource budgets.
What you need
Install a CTranslate2 configuration compatible with your accelerator and chosen precision. Separate the checkpoint conversion path from ordinary transcription in your setup notes.
What to try first
Run a folder of representative clips with consistent decoding settings. Measure transcription quality and throughput together, including the time to consume the returned segments.
Check a continuous recording and several short files. Threading and batching choices can favor one workload while offering little improvement for the other.
Where to look for failures
Libraries and accelerator requirements depend on the compute path. Changing beam size or silence processing can make performance comparisons misleading.
Related Franklin pages
Use the same source audio and decoding assumptions for transcription tests. Measure performance and listen to the mistakes, especially around silence, names and language changes.
- Recognize speech and explore emotion or audio-event annotations supported by the selected model. SenseVoice for multilingual speech and audio events
- Transcribe or translate speech with OpenAI's published Whisper models. Whisper for multilingual speech recognition
For an application where the quality of recording and handling audio matters, see Franklin’s Open Mic AI profile. Open Mic AI
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗