Transcriptinator

Cherry Blossom Development

About this app

Transcriptinator records or imports audio, transcribes it, identifies who's speaking, and — optionally — helps you turn what was said into actionable notes. Everything sensitive stays on your
device.

On-device from start to finish. Audio never leaves your phone. Transcription runs on open-source Whisper models locally. Speaker identification uses pyannote diarization and 3D-Speaker embedding
models — also local. You can use Transcriptinator on airplane mode; the only step that needs the network is the one-time model download.

Learns voices, not just sounds. Name a speaker once — "that's my colleague Dana" — and Transcriptinator auto-labels her in every future recording where she talks. Voice profiles are stored only on
your device and never uploaded.

Pick your quality tier. Three transcription tiers are available: Fast (Whisper base, ~200 MB), Balanced (Whisper small, ~600 MB), and Accurate (Whisper medium, ~1.5 GB). Set a global preference, or
retranscribe any single recording with a different tier in one tap.

Author your own rules. Turn spoken phrases into structured extraction cards you can act on later. Say "pet the cat" and get a todo. Say "let's meet Tuesday at 3" and get a calendar event. Two
matching modes:

- Offline (literal phrases) — deterministic, no API key, runs fully on-device.
- AI (semantic descriptions) — with an LLM configured, plain-English descriptions catch paraphrases automatically.

Optional AI, bring your own key. For document classification, summaries, and smart extraction of todos, calendar events, reminders, decisions, questions, and notes, connect one of three providers:

- Anthropic (Claude)
- OpenAI (ChatGPT)
- Google (Gemini)

You bring the API key. It's stored securely on your device — never on our servers, never in backups. Consent is opt-in and revocable at any time. Only text transcripts are ever sent to your provider
— never audio, never speaker embeddings. Monthly usage totals and per-recording token counts are shown right in the app so you always know what you're spending.

Every extraction cites its source. Tap a todo, calendar event, or reminder card to jump to the exact moment in the audio where it was said. Extractions carry the verbatim quote and link back to the
source segment so you can verify before acting.

What Transcriptinator is good at

- Meeting notes with 2–6 speakers
- Long-form interviews, lectures, sermons
- Voice memos that need real transcripts, not just recordings
- Field recorder imports (WAV including IMA ADPCM, MP3, M4A, OGG, FLAC)
- Building your own workflows with custom rules

What it doesn't do

- Live captioning during a call — everything is processed after recording
- Speaker verification for security — voice matching is a labelling convenience, not authentication
- Cloud sync — deliberately. Your recordings stay on your phone

Free, no ads, no upsells. Transcriptinator is free and will stay free. No ads, no in-app purchases, no premium tier. The AI features are genuinely optional — the app works fully without any external
service.

Publisher: Cherry Blossom Development, LLC

What’s new

• Copy transcripts to the clipboard in Plain text, Markdown, JSON, or CSV
• Save transcripts as .txt / .csv / .srt / .vtt via the system file picker
• Send a Markdown transcript straight to Obsidian (Advanced URI community plugin required)
• Link a speaker in People to a phone contact via the OS contact picker
• Two-tier recording menu — Share / Copy / Save to file / Send to Obsidian