Whisper Skill
Local Whisper transcription that picks the backend and model for your hardware
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 28, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Jul 30, 2026 · df6c19b
⚠ This skill is no longer available upstream. Our re-check on Aug 10, 2026 couldn't find it any more (repo unreachable/deleted). The test below is what we measured on Jul 28, 2026 and we're leaving it up as a record — but there is nothing left to install, so we've removed the command.
Transcribed the same 10-second clip two ways on an M3 Mac: the naive no-skill path (pip install openai-whisper, whisper sample.wav --model small) ran 5m44s on CPU and returned 'That week Brown Fox jumped over the lazy dog, whispered transcription benchmark's major word error rate' - 6 wrong words out of 26, about 23% WER. Following the skill, its detector correctly read the machine as Apple M3 / 8 GB / Metal and routed to mlx-whisper on MPS, and its example script returned a word-perfect transcript with clean SRT cues in 3m35s including the model download. One real bug: examples/common.py builds the MLX repo id as mlx-community/whisper-<model>, so the model its own detector recommends (--model small) dies on a 404, and you have to pass mlx-community/whisper-small-mlx by hand. Three links in the body (docs/word-level-subs.md, methodology/post-processing.md, examples/burn_subs.py) point at files that were never committed.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 5/5
- Output vs. baseline 8/10
- Docs & honesty 3/5
What Whisper Skill does
Sets up and runs Whisper locally instead of through the OpenAI API: it detects your OS, CPU, GPU and RAM, then picks a backend (mlx-whisper, faster-whisper, whisper.cpp, whisperx, openvino) and a model to match. Ships runnable examples for single-file and batch transcription, URL pull via yt-dlp, speaker diarization, push-to-talk voice dictation, and burning CapCut-style subtitles into MP4. Triggers on requests to transcribe audio or video locally, generate subtitles for shorts and reels, do speech-to-text without an API key, or replace Superwhisper-style dictation tools.
How to install Whisper Skill
Nothing to install: the source repository no longer has this skill. If the author brings it back, our daily re-check will pick it up and the command will reappear here.
Commands — how to trigger Whisper Skill
-
/whisper-skillLocal Whisper transcription that picks the backend and model for your hardware
It also activates on plain-language prompts like these:
-
Transcribe this podcast episode locally without using the OpenAI API -
Burn TikTok-style subtitles into my reel from this MP4 clip -
I want a free Superwhisper-style dictation tool for my Mac
Frequently asked questions
- Is the Whisper Skill skill free?
- Yes. The skill itself is free from Mobiss11/Whisper-Skill. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Whisper Skill work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 28, 2026. Verdict: Tested · Works. Transcribed the same 10-second clip two ways on an M3 Mac: the naive no-skill path (pip install openai-whisper, whisper sample.wav --model small) ran 5m44s on CPU and returned 'That week Brown Fox jumped over the lazy dog, whispered transcription benchmark's major word error rate' - 6 wrong words out of 26, about 23% WER. Following the skill, its detector correctly read the machine as Apple M3 / 8 GB / Metal and routed to mlx-whisper on MPS, and its example script returned a word-perfect transcript with clean SRT cues in 3m35s including the model download. One real bug: examples/common.py builds the MLX repo id as mlx-community/whisper-<model>, so the model its own detector recommends (--model small) dies on a 404, and you have to pass mlx-community/whisper-small-mlx by hand. Three links in the body (docs/word-level-subs.md, methodology/post-processing.md, examples/burn_subs.py) point at files that were never committed.
- What is the Whisper Skill SkillProof Score?
- 8.0/10 — installs cleanly 4/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 3/5.
- How do I install Whisper Skill?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Whisper Skill with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.