Overview & What It Does
ScribeCrate packages faster-whisper as a Docker transcription server with two familiar endpoints: /v1/audio/transcriptions and /v1/audio/translations, so any client written for the OpenAI Whisper API can point at your own machine. Drop in an MP3, M4A, WAV, WebM, OGG or FLAC file and get plain text, JSON, verbose JSON, SRT or WebVTT back; translation outputs English. Pick a Whisper model from tiny to large-v3 or large-v3-turbo. Optional local speaker diarization with sherpa-onnx labels who said what in interviews and meetings, and uploaded files can stream decoded segments over SSE. With models cached and WHISPER_LOCAL_ONLY enabled it runs fully offline. One docker run command starts it on port 9000 bound to localhost, and a fresh install generates its own API key. Written by Lin Song of Self-Hosted AI Stack, whose repos have over 40,000 GitHub stars. It handles files, not live microphone audio, and the sister project ScribeCrate Live covers that. Listed on theresanaiforthat.com as a release from about a month ago, rated 4.6 by 10 reviewers; the star count and the TAAFT rating were both checked on Nov 11, 2026.
A self-hosted transcription API with OpenAI-compatible endpoints: audio never leaves your server. Tested thoroughly for developer reliability, workflow friction, and value relative to competing AI stacks.
Pros & Cons: What We Like vs Considerations
✓ What We Like
- OpenAI-compatible endpoints mean existing Whisper API clients cover the common transcription calls by changing one URL
- MIT licensed with automated public GitHub Actions image builds, so the Docker image you pull is traceable to the source
- Runs offline once models are cached, on CPU amd64 or arm64, or on an NVIDIA GPU with the CUDA image
- Outputs SRT and WebVTT subtitles plus optional sherpa-onnx speaker labels at no license cost
✕ Things to Consider
- No hosted version: you need a Linux host, Docker and your own hardware budget before anything transcribes
- Handles uploaded files only; live microphone or stream audio needs the separate ScribeCrate Live project
- Speaker diarization is disabled in SSE streaming mode, so you choose between live segments and speaker labels
- It does not implement every parameter of the OpenAI audio API, so edge-case options may be missing
Pricing & Value Analysis
Pricing Tier: Free
No software fee: open source under the MIT license, verified Nov 11, 2026 on scribecrate.com and its GitHub repo (hwdsl2/scribecrate). You pay only your own server. The default setup runs on CPU with at least 700 MB free RAM for the base model; an NVIDIA CUDA image is available for amd64. There is no hosted plan and no usage-based billing on the official site, and no OpenAI account or key is needed because the API is self-hosted.
Best Alternatives to ScribeCrate
If ScribeCrate doesn't fit your budget or technical stack, consider these top-ranked alternatives in the same domain:
| Alternative Tool | Pricing | Best For | Rating | Compare |
|---|---|---|---|---|
| ElevenLabs | Freemium | YouTubers, podcasters, video game developers, e-learning creators, and international video localization. | ★ 4.9 | View Review → |
| Fireflies.ai | Freemium | Sales teams, remote companies, customer success managers, and agency project directors. | ★ 4.7 | View Review → |
| OpenAI Whisper | Free | Developers building transcription apps, video subtitlers, and privacy-conscious users running speech-to-text locally. | ★ 4.9 | View Review → |
Recommended Similar Tools
ElevenLabs
Voice & AudioThe gold standard for lifelike AI voice synthesis, voice cloning, and multilingual dubbing.
Fireflies.ai
Voice & AudioAutomated AI meeting assistant that joins video calls, transcribes discussions, and extracts action items.
OpenAI Whisper
Voice & AudioState-of-the-art open-source speech recognition supporting 99+ languages with robust background noise resilience.