whisper.cpp at a glance
Checked 15 Oct 2026whisper.cpp runs OpenAI's Whisper models in plain C/C++ with CPU-only inference plus Metal, CUDA, Vulkan, ROCm and other backends. Memory ranges from ~273 MB (tiny) to ~3.9 GB (large). Code is MIT and the ggml model conversions are tagged MIT. You build it with CMake; the CLI wants 16-bit WAV input. Experimental speaker-turn marking (tinydiarize) works with one English model only. It does not summarise meetings or join calls.
RepoQuarry. “whisper.cpp.” Checked 15 Oct 2026. https://repoquarry.dev/projects/whisper-cppHow much of ElevenLabs does it replace?
1 of 4 jobs. Local transcription; speaker turns only via experimental tinydiarize with an English small model.
For speech generation, voice cloning, voice agents, see other ElevenLabs alternatives. You may also need to build an interface, an API server and hosting yourself.
How much of Otter.ai does it replace?
1 of 3 jobs. Transcribes recordings locally; no live meeting capture, named speakers or summaries.
For speaker identification, meeting summaries, see other Otter.ai alternatives. You may also need to build an interface, an API server and hosting yourself.
Facts, with sources
Every answer links to where we found it. We read the documentation; we did not install it.
Support for CPU-only inferencegithub.com · whisper.cpp#readme ↗ 1d ago
license: mithuggingface.co · whisper.cpp ↗ 1d ago
download one of the Whisper models converted in ggml formatgithub.com · whisper.cpp#quick-start ↗ 1d ago
tiny | 75 MiB | ~273 MB ... large | 2.9 GiB | ~3.9 GBgithub.com · whisper.cpp#memory-usage ↗ 1d ago
Supported platforms: Mac OS (Intel and Arm) ... Linux / FreeBSD ... Windows (MSVC and MinGW)github.com · whisper.cpp#readme ↗ 1d ago
cmake --build build -j --config Releasegithub.com · whisper.cpp#quick-start ↗ 1d ago