whisper.cpp

What is whisper.cpp?

whisper.cpp runs OpenAI's Whisper speech-to-text models on ordinary CPUs (x86, ARM, Apple Silicon) and optionally on NVIDIA, AMD/Intel (Vulkan) or MUSA GPUs. Besides the command-line tool it includes whisper-server, an HTTP server with an OpenAI-like transcription API, and official Docker images that bundle ffmpeg.

Where it fits in a 10–200 person business

• Transcribe meeting, interview or call recordings on your own server instead of a cloud API
• Speech-to-text back-end for an internal assistant or automation workflow
• Offline subtitles for training or marketing videos

These use cases are our editorial suggestions, not claims made by the project.

Can a business use whisper.cpp? (license)

This is a plain-English summary of the license text, not legal advice. Third-party models you run with it have their own licenses.

Data & compliance: does whisper.cpp phone home?

Checked against the project's own docs and repository on Oct 22, 2026. "Not documented" means we could not find the answer in official sources; it does not mean the feature is missing or that no data is sent. Third-party services you connect (model APIs, vector databases, search engines) have their own data policies. Compare all apps.

What hardware does whisper.cpp need?

The project does not publish a minimum CPU/RAM figure, so we leave the official value blank rather than guess.

GPU: Optional. The default Docker image runs on CPU; CUDA, Vulkan and MUSA images are available.

No official server minimum. Official memory use per model is about 273 MB (tiny), 388 MB (base), 852 MB (small), 2.1 GB (medium) and 3.9 GB (large).

Source: github.com/ggml-org/whisper.cpp#memory-usage

How much does it cost to host whisper.cpp per month?

How to deploy whisper.cpp

Method: Docker

MODEL_PATH=/srv/whisper-models
mkdir -p "$MODEL_PATH"
docker run -it --rm -v $MODEL_PATH:/models ghcr.io/ggml-org/whisper.cpp:main \
  download-ggml-model.sh base /models
docker run -it --rm -p "8080:8080" -v $MODEL_PATH:/models ghcr.io/ggml-org/whisper.cpp:main \
  whisper-server --host 0.0.0.0 -m /models/ggml-base.bin
# POST audio to http://127.0.0.1:8080/inference

From the official instructions: github.com/ggml-org/whisper.cpp#docker. Always check the source for the current version before you run it.

Known gotchas when self-hosting whisper.cpp

• The project warns not to run whisper-server with administrative privileges and to keep it sandboxed, because it accepts file uploads and can call ffmpeg on them. Validate inputs and do not expose it publicly as is. source
• The server expects WAV input. Other formats need the --convert flag, which requires ffmpeg on the server (included in the official Docker images). source
• Models are not bundled. Download them first (download-ggml-model.sh) into a mounted folder, or every new container starts without a model. source

Alternatives in Speech & translation

• PDFMathTranslate PDF translator that keeps the original layout, formulas, charts and tables.
• LibreTranslate Self-hosted machine translation API with no dependency on Google or DeepL.
• Speaches OpenAI-compatible speech-to-text and text-to-speech server (faster-whisper, Kokoro, Piper).

Sources

• GitHub repository: github.com/ggml-org/whisper.cpp
• Official documentation, Deploy instructions, Gotcha 3: github.com/ggml-org/whisper.cpp#docker
• License text: github.com/ggml-org/whisper.cpp/blob/master/LICENSE
• Hardware requirements: github.com/ggml-org/whisper.cpp#memory-usage
• Gotcha 1, Gotcha 2: github.com/ggml-org/whisper.cpp/blob/master/examples/server/README.md
• Hetzner Cloud pricing (cost estimate): www.hetzner.com/cloud/regular-performance
• GitHub stars, commits and releases: GitHub REST API, fetched Oct 22, 2026