Privacy Policy

This policy explains what Speech Relay collects, why, how long we keep it, and how you control it.

What we collect

• Your account: when you sign in with GitHub or Google, we receive your email address, your name and an account identifier from that provider. A cookie keeps you signed in.
• Your sessions: the audio you stream and the transcript we produce.
• Usage records: how many seconds of audio each account and project processed. These contain no audio and no transcript text.
• Security records: operational logs and metrics to run and protect the service. To limit abuse we keep a salted hash of the caller's network address for one hour; we do not keep the address itself.
• The API waitlist: if you join, your name, email address and, if you give one, your organization. We use them only to contact you when API access opens.
• Website statistics: when you visit our websites, including the playground and your dashboard, Vercel counts the visit and records the page, the site that linked you, your country, and your browser and device type. It sets no cookies, and these statistics do not identify you. Page addresses that contain a session identifier are shortened before they are recorded.

Stored sessions

• Playground: saving is on by default. Audio and transcript are kept in your account for 30 days, then deleted automatically. You can choose 7, 30 or 90 days, or turn saving off, in your dashboard at any time.
• API: saving is off unless you turn it on for a project.
• You can listen to, download and delete any saved session yourself. A deleted or expired session is erased from our storage within 10 days.
• When saving is off, audio and transcript are used only to run the session and are not kept.

Other people in your recording

A recording can include people other than you. Before you record someone, tell them that the session is transcribed and, if saving is on, stored, and get their consent where the law requires it. If you cannot, turn saving off first.

Voice data

We do not create voiceprints or other biometric identifiers from your audio, we do not use it to identify anyone, and we never use your audio or transcripts to train models. Enrollment audio for target-speaker mode is used for that session only.

Who processes your data

• Microsoft Azure: runs the speech service and stores saved sessions, in its Central India region.
• Neon: hosts our account database, including saved transcripts.
• Vercel: hosts our websites and account app; the account app runs in Singapore.
• Cloudflare: network, DNS and email routing.
• GitHub and Google: sign-in.

Your data may also be processed in other countries where these providers operate. We do not offer a choice of region. We will update this list before a new provider starts handling your content.

Your choices and requests

You can change retention, download your sessions, delete sessions and delete your account from the dashboard. Deleting your account deletes your saved sessions and account details; usage records are kept for billing. For anything else, including access, correction or deletion requests, write to privacy@speechrelay.online.

Security

Data is encrypted in transit. Saved sessions are private to your account, and download links expire after one minute. If a security incident affects your data, we will tell you without undue delay.

Age and changes

The service is not for anyone under 18. If we change this policy we will update the date above, and tell account holders of any material change before it takes effect. Questions: contact@speechrelay.online.