whisper.cpp
Transcribe speech locally with compatible Whisper model weights and a native inference engine.
Published by ggml.org
How to use it
Build or install the runtime for your machine, obtain compatible Whisper weights, and run the transcription tool on a local audio sample. Use the output options suited to your caption or text workflow.
Select a model size based on memory, latency, and transcription quality. Accelerated builds have platform requirements. Compare the transcript against your audio before relying on timestamps or automatic language detection.
- Hardware
- CPU, NVIDIA CUDA, Apple Metal, Apple Core ML, Vulkan
- Model formats
- Whisper GGML weights
- Operating systems
- Linux, Windows, macOS, Android, iOS
- Access
- Runs locally. Supply compatible Whisper weights and local audio.