whisper-ctranslate2

Whisper command line client compatible with original OpenAI client based on CTranslate2.
It uses CTranslate2 and Faster-whisper Whisper implementation that is up to 4 times faster than openai/whisper for the same accuracy while using less memory.
Key features:
- ✅ 4x faster than OpenAI's Whisper at the same accuracy (up to 16X faster with batched inference)
- ✅ OpenAI Whisper CLI compatibility (easier migration)
- ✅ Docker image with pre-loaded models
- ✅ Speaker Diarization (Speaker Identification)
- ✅ Voice Activity Detection (VAD) Filter to improve quality
- ✅ Load your own fine-tuned Whisper models
- ✅ Live transcription from microphone
- ✅ Color-coded confidence visualization
Installation
Python package
To install the latest stable version, just type:
pip install whisper-ctranslate2
Using prebuild Docker image
You can use build docker image. First pull the image:
docker pull ghcr.io/softcatala/whisper-ctranslate2:latest
The Docker image includes the small, medium and large-v2 models.
To run it:
docker run --gpus "device=0" \
-v "$(pwd)":/srv/files/ \
-it ghcr.io/softcatala/whisper-ctranslate2:latest \
/srv/files/e2e-tests/gossos.mp3 \
--output_dir /srv/files/
Notes:
* _--gpus "device=0"_ gives access to the GPU. If you do not have a GPU, remove this.
* _"$(pwd)":/srv/files/_ maps your current directory to /srv/files/ inside the container
If you always need to use a model that is not in the image, you can create a derived Docker image with the model preloaded or use Docker volumes to persist and share the model files.
CPU and GPU support
GPU and CPU support is provided by CTranslate2.
It has compatibility with x86-64 and AArch64/ARM64 CPU and integrates multiple backends that are optimized for these platforms: Intel MKL, oneDNN, OpenBLAS, Ruy, and Apple Accelerate.
GPU execution requires the NVIDIA libraries cuBLAS for CUDA 12 and cuDNN 9 for CUDA 12 to be installed on the system. Please refer to the CTranslate2 documentation
By default the best hardware available is selected for inference. You can use the options --device and --device_index to control manually the selection.
Usage
Same command line as OpenAI Whisper.
To transcribe:
whisper-ctranslate2 inaguracio2011.mp3 --model medium
To translate:
whisper-ctranslate2 inaguracio2011.mp3 --model medium --task translate
Whisper translate task translates the transcription from the source language to English (the only target language supported).
Additionally using:
whisper-ctranslate2 --help
All the supported options with their help are shown.
CTranslate2 specific options
On top of the OpenAI Whisper command line options, there are some specific options provided by CTranslate2 or whiper-ctranslate2.
Batched inference
Batched inference transcribes each segment independently which can provide an additional 2x-4x speed increase:
whisper-ctranslate2 inaguracio2011.mp3 --batched True
You can additionally use the --batch_size to specify the maximum number of parallel requests to model for decoding.
Batched inference uses Voice Activity Detection (VAD) filter and ignores the following parameters: compression_ratio_threshold, logprob_threshold,
no_speech_threshold, condition_on_previous_text, prompt_reset_on_temperature, prefix, hallucination_silence_threshold.