whisper-voice
Skill16 starsChanged 9 months agoMITView on GitHub
Native macOS menu bar app for live voice-to-text with auto-type using WhisperKit on Apple Silicon
In this repository39 other files it gives its agents.
- add-webhookSkill
- aws-production-deploySkill
- build-in-public-youtubeSkill
- casualize-namesSkill
- classify-leadsSkill
- create-proposalSkill
- cross-niche-outliersSkill
- design-websiteSkill
- euron-qaSkill
- excalidraw-diagramSkill
- excalidraw-visualsSkill
- generate-reportSkill
- ghost-browserSkill
- gmail-inboxSkill
- gmail-labelSkill
- gmaps-leadsSkill
- image-to-videoSkill
- instantly-autoreplySkill
- instantly-campaignsSkill
- linkedin-carousel-handwrittenSkill
- literature-researchSkill
- local-serverSkill
- mac-controlSkill
- modal-deploySkill
- Nano Banana 2 Image Generation MasterSkill
- onboarding-kickoffSkill
- pan-3d-transitionSkill
- recreate-thumbnailsSkill
- scrape-leadsSkill
- send-telegramSkill
- skill-builderSkill
- skool-monitorSkill
- skool-ragSkill
- thumbnail-generatorSkill
- title-variantsSkill
- upwork-applySkill
- video-editSkill
- welcome-emailSkill
- youtube-outliersSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
Wait for model download (~140MB for base model)
- Distinctive: line 62
No API keys needed (fully offline)
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Whisper Voice — Live Speech-to-Text Mac App
Goal
Build and run a native macOS menu bar app that captures live microphone audio, transcribes it offline using WhisperKit (on Apple Silicon), and auto-types the text wherever the cursor is.
Inputs
| Name | Type | Required | Description |
| model_size | string | No | Whisper model: tiny, base (default), small |
| language | string | No | "en" (default) or "hi" for Hindi mode |
| chunk_duration | float | No | Seconds per audio chunk (default: 3.0) |
Process
1. Build the app
cd AiwithDhruv_Voice/WhisperAiwithDhruv
swift build2. Run the app
swift run WhisperAiwithDhruv
# Or open in Xcode: open Package.swift → Cmd+R3. First launch setup
- Grant microphone permission when prompted
- Grant Accessibility in System Settings → Privacy → Accessibility
- Wait for model download (~140MB for base model)
4. Usage
- Cmd+Shift+Space — Toggle recording on/off
- Click mic icon in menu bar for controls
- Speak — text auto-types at cursor position
- Toggle Hindi mode for Hindi/Hinglish input
Outputs
| Name | Type | Description |
| transcribed_text | string | Live transcribed text typed at cursor |
| history | array | Last 50 transcription entries in menu bar |
Edge Cases
- No mic: Shows error in menu bar dropdown
- Accessibility denied: Auto-type disabled, manual copy from history
- Silence: VAD skips silent chunks (energy-based threshold)
- Hallucinations: Filters common Whisper artifacts ("Thank you.", "...")
- Model not downloaded: Shows download progress bar
Environment
- macOS 14+ (Sonoma)
- Apple Silicon (M1/M2/M3/M4)
- Xcode 15+ (for building)
- No API keys needed (fully offline)
Schema
Inputs
| Name | Type | Required | Description |
| model_size | string | No | tiny / base / small |
| language | string | No | en / hi |
| chunk_duration | float | No | 2.0 - 8.0 seconds |
| silence_threshold | float | No | 0.002 - 0.05 |
Outputs
| Name | Type | Description |
| transcription | string | Live text output |
| auto_typed | boolean | Whether text was injected at cursor |
Credentials
| Name | Source |
| None | Fully offline, no API keys |
Composable With
video-edit (add transcription captions), send-telegram (send transcriptions to phone)
Cost
Free — runs entirely on-device. Model download is one-time (~140MB for base).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.