Turn rambling into action, in real time.

Brain dump out loud while you work. Good Ramble listens and quietly writes back next steps, notes, and answers.

Start rambling

Bring your own Deepgram & OpenAI keys. No account.

Animated demo: you speak, Good Ramble transcribes what you say live, and once you pause it replies silently in text with suggestions and code.

It knows when to say something.

Your instructions decide what deserves a reply. Filler and half-finished thoughts get silence. Questions, mistakes, and whatever you asked it to watch for get an answer.

Your instructions: Help me debug this out loud. Flag edge cases I'm missing and keep a running list of TODOs.

  • okay, um… let me pull up the cache module

    Stays quiet

  • so writes go straight to localStorage — wait, what if two tabs write at once?

    Replies: That's a race: last write wins silently. Add a version check before the write.
  • yeah, let me just rename this to something clearer

    Stays quiet

  • okay, I think the fix is done

    Replies: TODO: test with two tabs open, and handle the rejected write in the UI.
  1. Step 1

    Tell it what matters

    Write custom instructions or pick a template — rubber duck debugging, interview prep, and more.

  2. Step 2

    Think out loud

    Talk the way you normally would. It transcribes in real time.

  3. Step 3

    Get replies worth reading

    When you pause, it checks your instructions and only writes back if it's useful. Keep talking to interrupt.

For any kind of talking-to-yourself

You say

...so the cache invalidation happens when — wait, what if two tabs write at once?

It writes back

  • Two tabs writing to the same key is a race, not an edge case.
  • if (cached.version !== expected) return
  • Worth deciding: last-write-wins, or reject and re-fetch?

Your ramble stays yours.

No history stored on our server. Everything is stored locally in your browser.

Your browser

API keysChatsRecordingsTemplates

Deepgram

Speech-to-text, in realtime

OpenAI

Generates the reply

Good Ramble server

— nothing stored here
  • No account or sign-up.
  • Keys stay in localStorage and are sent only to the providers you choose.
  • Chats and recordings live in IndexedDB, on your device.
  • Export transcripts and recordings, or delete chats, any time.

Note: audio and text you send do reach Deepgram and OpenAI directly from your browser — we just don't keep a copy on a server.

Features

Multi-speaker mode

Transcribe a conversation, not just one voice — each speaker is tracked separately.

Speaker 1Speaker 2

Recordings & transcripts

Download the audio and the full transcript whenever you want.

mp3

Attach files

Drop in files for context while you talk.

spec.pdf

MCP tools

Connect MCP servers so it can act on what you're saying.

search_docs()

Templates

Start from a ready-made prompt instead of writing your own.

Rubber duckInterview prepDecisionMeeting notes

Usage & cost tracking

See what you're spending on Deepgram and OpenAI, at a glance.

Frequently asked

Do I need an account?

No. There's no sign-up — open the app and start talking.

What does it cost?

Good Ramble itself is free. You bring your own Deepgram and OpenAI API keys and pay those providers directly for what you use. The usage panel tracks spend as you go.

Where is my data stored?

Locally, in your browser. Chats, recordings, and templates live in IndexedDB on your device — there's no server history.

Can it talk back out loud?

No — replies are text only, on purpose, so it never interrupts you mid-thought.

Does it work with multiple people?

Yes — multi-speaker mode transcribes a conversation, not just one voice.

Go on, ramble.

Start rambling