obra/ghost-pepper
Hold-to-talk speech-to-text for macOS. 100% local, powered by WhisperKit and local LLM cleanup. Hold Control to record, release to transcribe and paste.
https://skillcdn.ai/gh/obra/ghost-pepper내 AI에 이 주소를 연결하면 이 스킬들을 쓸 수 있어요. 연결 방법 보기
- 미확인
- 기본 브랜치
main - 커밋
a9eed76 - 라이선스
선언되지 않음
저장소 둘러보기
스킬과 문서의 원문에서 찾습니다. 화면용 번역은 검색되지 않아요.
문서
- Onboarding Flow DesignFirst-run onboarding wizard for GhostPepper. Shows once on first launch, walks the user through permissions, model download, and a live "try it" step. Persists completion to UserDefaults so it never…docs/superpowers/specs/2026-03-20-onboarding-design.md
- Transcription Stack DesignAdd first-class configurable recording chords, cleanup backend/model selection, OCR-informed cleanup, and a persistent correction system without turning the current app into a single AppState blob.docs/superpowers/specs/2026-03-23-transcription-stack-design.md
- Cleanup Model Probe DesignAdd a small local command-line probe for Ghost Pepper's cleanup models so Bot can reproduce and inspect failures like the Qwen 3 fast-model reasoning leak with exact inputs.docs/superpowers/specs/2026-03-25-cleanup-model-probe-design.md
- FluidAudio STT Backend DesignGoal: Add a second local speech-to-text backend so GhostPepper can offer Whisper and Parakeet-family transcription models from one model picker without exposing backend details in the UI.docs/superpowers/specs/2026-03-25-fluid-audio-stt-design.md
- Settings Sidebar And Sound Toggle DesignDate: 2026-03-25docs/superpowers/specs/2026-03-25-settings-sidebar-and-sound-toggle-design.md
- Transcription Lab DesignAdd a user-facing Transcription Lab section to Settings that keeps the last 50 completed recordings with non-empty audio and lets users rerun those recordings with different speech models, cleanup mo…docs/superpowers/specs/2026-03-25-transcription-lab-design.md
- FluidAudio Diarization DesignAdd a first-pass Ignore other speakers feature that works only on the FluidAudio speech path, uses diarization to follow the first substantial speaker in a recording, and improves stop-and-paste dict…docs/superpowers/specs/2026-03-26-fluidaudio-diarization-design.md
- Speaker Identities DesignAdd durable speaker identities on top of transcription-lab speaker tagging so Jesse can:docs/superpowers/specs/2026-04-20-speaker-identities-design.md
- Agentic Meeting Q&A — DesignStatus: Draft for review Author: Claude Code (Opus 4.7) brainstorming session, 2026-04-28 Companion spec: 2026-04-28-cross-meeting-qa-claude-api-design.md (predecessor — single-shot cram-mode Claude…docs/superpowers/specs/2026-04-28-agentic-meeting-qa-design.md
- Cross-Meeting Q&A: Claude API Backend, Streaming, Cost DisplayCommit 1a39495 ("feat: cross-meeting Q&A with Qwen3 8B support") forces the Qwen 3 8B local model (~5 GB, 8B params) for every cross-meeting Q&A request. On Matt's hardware this freezes the machine.…docs/superpowers/specs/2026-04-28-cross-meeting-qa-claude-api-design.md
- Meeting Window: Today's Calendar List + currentMeeting() HardeningTwo related issues:docs/superpowers/specs/2026-04-28-meeting-home-calendar-list-design.md