whisper-cpp mit Homebrew, dnf, Nix, apk, scoop installieren

Projektlinks

• Homepage
• Upstream-Dokumentation
• Paketmanager-Seite

Befehle und Aliase

• whisper-bench
• whisper-cli
• whisper-command
• whisper-lsp
• whisper-quantize
• whisper-server
• whisper-stream
• whisper-talk-llama
• whisper-vad-speech-segments

Projektgeschichte

OpenAI introduced Whisper on September 21, 2022 as an automatic speech recognition system trained on 680,000 hours of multilingual and multitask supervised data. The original OpenAI repository positioned Whisper as a general-purpose model for multilingual transcription, translation to English, and language identification.

The whisper.cpp repository was created on September 25, 2022. Its README describes the project as a port of OpenAI's Whisper model in C/C++, with the high-level model implementation contained in `whisper.h` and `whisper.cpp` and the lower-level tensor work handled by ggml. That made Whisper usable in environments where Python, PyTorch, or CUDA were undesirable or unavailable.

The project grew with ggml-style model conversion, quantization, platform backends, examples, bindings, and command-line tools. Its model documentation explains that original OpenAI PyTorch models are converted to ggml format for loading from C/C++, while project discussions and release notes document features such as integer quantization, streaming, server usage, and embedded-device experiments.

Adoptionsgeschichte

whisper.cpp became one of the canonical local-AI command-line packages because it made a large neural speech model feel like a normal Unix tool: download a converted model, run an executable, and transcribe local audio without a cloud service. The upstream GitHub repository reported more than 51,000 stars and more than 5,700 forks via GitHub's API on July 2, 2026.

Its adoption also matters historically because it helped establish the ggml pattern later associated with local inference projects: compact C/C++ runtimes, quantized model files, CPU-first portability, and optional acceleration paths. In package-manager culture, whisper.cpp is the package people reach for when they want OpenAI Whisper behavior in scripts, media pipelines, offline transcription jobs, and low-power devices.

Wie es verwendet wird

Typical usage is local speech-to-text: converting or downloading a ggml-format Whisper model, running `whisper-cli` or related tools against audio, and emitting transcripts or timestamps. The repository also ships examples and executables for benchmarking, streaming microphone input, running a local server, quantizing models, and integrating with other applications through a C API or bindings.

The package is especially useful when privacy, offline operation, repeatable batch processing, or hardware portability matter more than managed cloud transcription. Users choose model size and quantization level as the tradeoff knob between accuracy, memory footprint, and speed.

Warum Paket-Nerds sich dafür interessieren

For package nerds, whisper.cpp is a landmark local-inference package: it compressed a research model into a buildable C/C++ artifact that could live comfortably in Homebrew, Linux distributions, Docker images, mobile apps, WebAssembly experiments, and single-purpose transcription workflows.

It also became a reference point for later local-AI ports. When people compare local speech runtimes, they often distinguish broad cross-platform portability in whisper.cpp from Apple-specific Core ML and Neural Engine approaches such as WhisperKit, or Python/GPU-oriented projects such as faster-whisper.

Zeitleiste

• 2022-09-21: OpenAI introduced Whisper and released model code and weights for multilingual speech recognition.
• 2022-09-25: The whisper.cpp GitHub repository was created.
• 2023: whisper.cpp discussions and releases documented integer quantization, streaming, and embedded-device experiments.
• 2026-07-02: GitHub API metadata showed the repository above 51,000 stars, reflecting broad adoption in the local-AI ecosystem.

Related projects

• OpenAI Whisper is the upstream model and Python reference implementation.
• ggml is the low-level machine-learning library and model-format ecosystem used by whisper.cpp.
• WhisperKit targets Apple Silicon with Swift and Core ML rather than whisper.cpp's cross-platform C/C++ approach.

Quellen

• OpenAI Whisper announcement: https://openai.com/index/whisper/
• OpenAI Whisper repository: https://github.com/openai/whisper
• whisper.cpp GitHub API metadata: https://api.github.com/repos/ggml-org/whisper.cpp
• whisper.cpp model documentation: https://github.com/ggml-org/whisper.cpp/blob/master/models/README.md
• whisper.cpp repository and README: https://github.com/ggml-org/whisper.cpp
• whisper.cpp v1.4.0 quantization discussion: https://github.com/ggml-org/whisper.cpp/discussions/838