Frequently Asked Questions

Dictate is a WebSocket relay for real-time speech-to-text. Your client opens a WebSocket connection, streams audio frames to it, and receives incremental recognition text back. Dictate forwards the frames to an upstream streaming speech recognition model and meters usage by audio duration. Your application never has to hold or rotate an upstream API key.

Three steps: create an account, request a short-lived WebSocket token from the /api/v1/ws-token endpoint, then open a WebSocket to /ws with that token and start sending audio frames. Recognition results arrive as incremental text on the same connection.

No. The Free tier gives you 30 minutes of transcription and needs no credit card. You only add a payment method when you upgrade to Pro or Enterprise.

The Free plan is the trial: 30 minutes of transcription and 100 API calls with no expiry date. It is a permanent free tier rather than a time-boxed trial.

16 kHz, 16-bit, mono PCM audio. WAV, MP3 and OGG containers are also accepted. Send frames of 100-200 ms: smaller frames add overhead without improving latency, and larger frames make results feel less live.

As binary frames, one frame per 100-200 ms of audio. Dictate relays frames frame-by-frame and never decodes, mixes, records or stores them.

The /ws path on the same host that serves the API, with the token as a query parameter: wss://your-host/ws?token=YOUR_TOKEN. The exact wss:// URL is returned in the token response so you never have to assemble it yourself.

5 minutes. Request a fresh token per session rather than caching one, which keeps a leaked token from staying useful for long.

By audio duration in milliseconds, measured server-side against the number and size of frames you actually send. Quota is decremented atomically per session, so concurrent sessions cannot overspend the same remaining quota.

Free: 30 minutes and 100 API calls. Pro at $9/month: 10 hours of transcription and 10,000 API calls. Enterprise at $49/month: 50 hours of transcription. Every plan resets at the start of each billing period.

New sessions stop being accepted once the included duration for the billing period is used. Nothing is charged beyond your plan price - quota is a limit, not a metered bill.

Open Profile, go to the Plan section and choose Cancel subscription. Access continues until the end of the period you already paid for, and you are not charged again.

Yes. The first payment carries a 7-day refund guarantee. Later renewals are refundable pro-rata if you contact support before the period ends.

Waffo is the default payment channel and Creem is retained as a fallback. The processor named on your statement depends on which channel handled that checkout. Either way, full card details never touch Dictate servers - the processor charges your card and issues refunds directly.

No. Audio frames are relayed frame-by-frame and are never decoded, recorded or written to disk. Dictate stores usage metadata only - session id, duration, frame count, sample rate and timestamps - because quota accounting needs it.

No. Neither your audio nor your transcripts are used for training, by us or as part of any arrangement with the recognition provider. They exist only for the duration of the live session.

The Volcano Engine Doubao streaming speech recognition model (bigmodel), reached through the Volcano Engine Open Speech Platform. Dictate relays to it and is an independent service, not an official product of the model provider.

Upstream keys never reach your client. They are held server-side and stored encrypted at rest, so a compromised browser cannot extract them.

Six categories: sexual or NSFW content, violence and gore, hate speech, child unsafe content (CSAM, zero tolerance), deepfakes and impersonation, and copyright or trademark infringement. See the Acceptable Use Policy for reporting channels and enforcement.

Check three things in order: the token has not expired (they last 5 minutes), the frames are 16 kHz 16-bit mono PCM, and the connection was opened as binary rather than text. The dashboard shows the remaining quota for the current period.

Email report@dictate.top. Reports of child safety issues are treated as urgent and triaged first. Urgent reports are acknowledged within 12 hours; other reports within 2 business days.

Email appeal@dictate.top with your account email and a short description of the action. Appeals are reviewed by someone who was not involved in the original decision.