Connect AssemblyAI for
diarized transcription.
Plug in your own AssemblyAI API key and Thought Recorder will return transcripts that label each turn with [Speaker A], [Speaker B], and so on. Billed to your account. We never see your audio or your key.
For interviews and meetings, knowing who said what changes everything.
Apple's on-device Speech framework and OpenAI's Whisper both produce one long block of text, regardless of how many people are talking. That's fine for solo voice notes — useless for interviews, podcasts, board meetings, or any recording where the value is in who said it as much as what they said.
AssemblyAI's speaker diarization separates the audio into turns and labels each one. Every paragraph in your transcript starts with the speaker who said it. Editors, journalists, podcasters, and anyone reviewing a multi-person conversation get hours of post-production time back.
The trade is small: about 65 cents per hour of audio, billed directly to your own AssemblyAI account. The connection is between your iPhone and AssemblyAI; Thought Recorder is just the courier.
Four steps. Ten minutes. Done.
-
01
Create an AssemblyAI account
Go to assemblyai.com/dashboard/signup. Sign up with your email (or Google / GitHub). You'll need to verify your email and a phone number — AssemblyAI's anti-abuse step, not anything we control.
New accounts get $10 in free trial credits — about 15 hours of audio. Plenty to test whether diarization fits your workflow before adding a card.
-
02
Add a payment method After trial
Once your free credits run out, go to Account → Billing and add a card. AssemblyAI is pay-as-you-go — billed monthly for what you actually use.
Tip: if you want a hard cap on spending, AssemblyAI lets you pre-pay credits instead of running a tab. Same billing page.
-
03
Copy your API key
Go to your API keys page. AssemblyAI generates one for you automatically when you sign up — you'll see a long string of letters and numbers under "Your API key". Click the copy icon.
Treat this like a password. Anyone with the key can use your AssemblyAI account. Don't paste it into chats, screenshots, or shared documents.
-
04
Paste it into Thought Recorder
Open Thought Recorder on your iPhone. Go to Settings → Diarized transcription (AssemblyAI). Paste your key into the field, wait a beat for the green checkmark, and you're done.
From now on, any recording can be transcribed with diarization by tapping the 🗣 AssemblyAI pill on the recording detail screen. On-device transcription still works alongside it — AssemblyAI is opt-in per file unless you flip the "Use AssemblyAI for new recordings" toggle in Settings.
$0.65 per hour of audio (Universal-2 model)
Pricing reflects AssemblyAI's Universal-2 API rate as of 28 May 2026. Check assemblyai.com/pricing for the current rate.
What goes where, plainly.
- Your API key lives in the iOS Keychain on your iPhone. It is not sent to Gonzodocs, Apple, or any third party — only to AssemblyAI when you transcribe.
- The audio file you transcribe is uploaded directly from your phone to AssemblyAI's API. We are the courier; we don't see the request, the response, or the audio.
- AssemblyAI's API terms state they do not use API request data to train their models by default, and they delete audio from their servers after processing. You can verify this at assemblyai.com/legal/privacy-policy.
- You can rotate or revoke your key anytime from the AssemblyAI dashboard. If you do, Thought Recorder will simply stop using AssemblyAI and fall back to on-device transcription.
Common questions.
Do I have to use AssemblyAI?
No. On-device transcription works for every recording without any key. AssemblyAI is an opt-in for files where speaker labels matter — interviews, meetings, panel discussions, podcasts.
Will I be charged just for connecting the key?
No. You're only charged per transcription request. Connecting the key has no cost, and new accounts come with $10 in free trial credits.
How accurate is the speaker labeling?
Very good for interview-style audio with clear turn-taking. Worse with overlapping speech, similar-sounding voices, or noisy environments. AssemblyAI uses generic labels (Speaker A, Speaker B) — they don't know who anyone is by name.
Can I rename "Speaker A" to "Alice"?
Not yet — for now, the label is part of the transcript text. You can use the transcript's Edit button to find-replace manually. A proper rename UI is on the roadmap.
What if I lose or accidentally share my key?
Go to your API keys page, revoke the old one, and generate a new one. Then paste the new key into Thought Recorder's Settings.
Can Thought Recorder use my key for other AssemblyAI services?
No. The app only calls AssemblyAI's transcription endpoint with speaker labels. It cannot run sentiment analysis, content moderation, or other features on your key.
Why not use a Gonzodocs-managed AssemblyAI key?
Because then we'd see your audio, manage your billing, and stand between you and AssemblyAI. Bring-your-own-key keeps us out of that loop entirely.
Whisper vs AssemblyAI — which should I pick?
Whisper for solo voice notes where you want the cheapest accurate transcription. AssemblyAI for any audio with two or more people where speaker labels matter. You can have both keys connected — Thought Recorder lets you pick the engine per recording.
Set up. Ready to go.
Every conversation, one tap from a labeled transcript.