Your data
WAV is a cloud service. When you record or upload a meeting, the audio leaves your device so we can transcribe it. This page says exactly where it goes, who touches it, and how to get rid of it.
The short version
Meeting audio is sent to the cloud. It is stored with Google Cloud and transcribed by Google’s speech service, with ElevenLabs as a backup.
AI notes see text, not audio. Titles, summaries, notes and answers to your questions are written by OpenAI from the transcript.
Voiceprints stay yours. The voice profiles that let WAV name speakers belong to your account only. They are never shared and never exported.
You can delete everything. Deleting a recording erases its audio, transcript and notes. Deleting your account erases all of it.
Who processes what
| Service | What it receives | Why |
|---|---|---|
| Google Cloud | Your audio, transcripts and notes (stored). App servers run in Google’s asia-east1 region. | Storage, sign-in (Firebase) and running WAV |
| Google Speech-to-Text (Chirp 3) | Meeting audio, processed in Google’s US speech region | Default transcription and speaker separation |
| ElevenLabs | Meeting audio, only when the default engine fails | Backup transcription |
| OpenAI | Transcript text (never audio) | Titles, summaries and notes, Ask answers, dictionary corrections, 書面語 conversion |
| Supabase | Your profile, plan usage, projects and voiceprints | Account database |
| Stripe | Billing details (no audio, no transcripts) | Payments |
| Resend | Your email address | Sign-in codes |
Dictation
Dictation in the Mac app is transcribed and tidied up by speech and language models that WAV runs itself. It isn’t sent to Google, ElevenLabs or OpenAI.
Training
Mixed-language recordings are where speech models improve most, so we ask before using recordings to train models:
- Trial accounts: on by default. You can switch it off for free in Settings, at any time. Switching off is never a paid feature.
- Paid plans: off unless you switch it on.
- EU, UK, Illinois, Texas and Washington: off unless you switch it on, on every plan.
Recordings capture other people. Only turn this on when everyone in the room has agreed.
Keeping and deleting
A recording: delete it from its ··· menu. WAV erases its audio, transcript, notes and earlier versions.
Your account: delete it in Settings › Profile. WAV erases your data from the product, and from the training archive, within 30 days, unless the law requires us to keep it. If you opted in to training, anything already built into a released model can’t be removed from that model.
A trimmed recording: the untrimmed original is kept for 30 days so you can undo the trim, then deleted.
Export: Settings › Profile › Export gives you a copy of your data.
Connected AI tools
If you connect Claude, ChatGPT, Cursor or a script to WAV with an API key, that tool can read your transcripts and notes when you ask it to. It can’t read your audio. Revoke a key in Connect AI and its access stops at once.
Questions
Read the Terms and the Privacy Policy when you create an account. If something here is unclear, ask us before you record.