The quote you wanted to keep, the decision that changed your mind, the one argument worth remembering — gone by next week, buried somewhere in 47 minutes of audio you'd have to scrub through again. vocateca turns everything you listen to into clean, searchable transcripts — entirely on your Mac — so nothing worth keeping slips away again.
Then you bring the intelligence — any LLM you trust, a knowledge base that stays yours.
Free On-device No account needed macOS 13+
Sound familiar?
You pour real hours into listening — and the best of it slips through your fingers. The notes you meant to take. The line you wanted to quote. The thread you'd swear you'd remember.
Podcasts on the commute, talks at 1.5×, interviews you keep meaning to revisit. Time in — and almost nothing you can search later.
That sharp point, that name, that number — clear in the moment, gone by the weekend. You remember that it mattered, not what it was.
Weeks later you want that one thing back — and it means scrubbing a timeline bar, guessing at minutes. Usually it's just lost.
It doesn't have to work that way. Every word, captured and searchable — on your Mac, the moment you press play on your own terms.
The honest workflow
Most tools blur the seam between transcription and “insight.” We don't. vocateca does two things exceptionally well — capture and clean — then hands you the raw material for whatever you want to think with.
Add a podcast, YouTube channel, or Instagram account. vocateca downloads and transcribes locally with the Parakeet-TDT engine — no cloud, no account, no audio ever leaving the machine.
Every transcript lands as tidy Markdown — push it to Obsidian or Notion, or pipe it anywhere with the CLI, webhooks, and MCP. Timestamps, speakers, and structure all intact.
Drop the transcripts into any model you trust to summarize, cross-reference, and query. Ask questions across a year of listening. vocateca never picks the model — you do.
Three steps, and the first two are done for you — on your Mac.
LLM-agnostic, by design
There is no shortage of tools that will pick a model for you, summarize behind a closed door, and quietly decide what your notes should say. We think that's the wrong trade. The transcript is the foundation — so we make it excellent, keep it open, and let you choose the intelligence that sits on top.
What vocateca does
What you decide
And how you build it is yours too — a Karpathy-style LLM wiki an assistant keeps evolving, a knowledge graph, a plain RAG index, or your own synthesis. We ship the transcript; the method is yours.
Get the clean transcripts. Bring your own intelligence.
The power of your inspiration stays in your hands.
We give you the transcript — clean, complete, and unopinionated — and get out of the way. No model imposed on what you think. No lock-in on the tools you use to think it.
What it does
The whole job of vocateca is to make the best possible transcript and keep it close at hand — so it's ready the moment you want to search it, export it, or reason over it with a model.
Podcasts, YouTube, and Instagram, transcribed locally with the Parakeet-TDT engine and Whisper as a fallback. Fast, accurate, and completely offline.
Every word of every transcript, instantly searchable across your whole archive — no server, no index in the cloud. Find the one thing someone said six months ago in seconds.
Clean Markdown straight into Obsidian or Notion, or wire it anywhere with webhooks, the CLI, and MCP — so transcripts become drop-in input for whatever you build next.
Follow a show once and keep an eye on the topics you care about. Set keywords and get told when a new episode mentions them — the library keeps watch so you don't have to.
Open core by heart
The transcription engine, the CLI, and the MCP server are open source. You can read exactly how your audio becomes text, extend it to fit your own workflow, and automate the whole thing from a script. Because the exits are open, your transcripts are never trapped — MCP makes them first-class input to AI tools, and the CLI pipes them straight into whatever comes next.
# the transcript is yours — pipe it wherever you think $ vocateca export --show "The Knowledge Project" --format md \ | your-llm --prompt "summarize the key decisions" → Ep. 214 · 47 min · exported (Markdown, 8.2 KB) → piped to your model of choice — vocateca picks none # or expose the whole library to an AI tool over MCP $ vocateca mcp --serve → transcripts now first-class input for any MCP-aware client
Private by default
Downloading and transcription run entirely on-device. No server sees your audio, your transcripts, or what you listen to, and there's no account to create just to use the app. Your library is a folder on your own disk.
The only time anything leaves your Mac is the moment you choose to — when you push a transcript to Notion, or hand one to a cloud model. That decision is always yours, and always explicit.
Pricing
Everything you need to produce transcripts and get them out — search, all sources, every export — is free, forever. Pro only automates the timing: it changes when the work happens, never what you get.
Everything you need to build the library by hand — and to feed any LLM you like.
On-device · no account needed · macOS 13+
Or €49/year. About a coffee a month — and far less than the hours you'd otherwise spend taking your own notes and hunting back for that one piece you wanted to keep. Same app; it just never waits for you to press a button.
Pro is automation, not a paywall on your transcripts. Nothing you produce is ever held back, watermarked, or sent somewhere for processing — Free and Pro make the exact same clean, exportable text.
Start free today. Automate it later — only if you want to.
Turn the hours you already listen to into clean transcripts on your Mac — then build a knowledge base with whatever model you trust. No lock-in. No account. No opinion imposed.
Free forever for manual use No account needed macOS 13+ · Apple silicon & Intel