capability
verified July 2026
Standby lines for the vendor’s bad night.
A standby failover line is capacity you hold but do not use: it runs beside your current TTS provider and activates automatically when that provider goes dark. It prices the insurance separately from the capacity — which is what makes the insurance worth holding.
01
How it works
- You keep your primary vendor. Standby lines sit behind our endpoint, integrated and tested, at $20 a month each.
- Activation is automatic when the primary stops answering — the failover path is exercised before the outage, not written during it.
- An activated standby serves at the same measured figures as any line: 107 ms first audio, single stream, on the production API.
const PRIMARY = "https://tts.your-current-vendor.example"
const STANDBY = "https://tts.gandr.ai" // integrated, tested
async function synthesize(req) {
try {
return await speak(PRIMARY, req)
} catch {
// same three surfaces, same request shape —
// the standby is the integration you already wrote
return await speak(STANDBY, req)
}
}
02
Why it is a separate product
Failover sized as a second full fleet doubles the voice bill against an event measured in hours a year, so most teams skip it and hope. A standby line prices the readiness itself — the socket that is integrated, tested, and waiting — apart from the talking, which makes holding the insurance rational. Our own serving practices the same discipline, drawn on the reliability sheet.
03
Notes — an engineer's checklist
01When does a standby line start billing as a full line?
It does not — a standby is $20 a month whether it activates or not. If you keep traffic on it permanently, that is a committed line and the honest move is to buy one at $150.
02How do I test the failover path without an outage?
Point a canary share of traffic at the standby base URL on a schedule — the request shape is identical, so the test is a config change. An untested failover path is a hope, not a plan.
03Does the standby clone the same voices?
Yes — the reference clip rides inside each request, so the same clip produces the same identity on the standby that it does anywhere else. No voice migration is involved in an activation.
See also
Related sheets.
capability
3 tiers
Serving that degrades before it fails
Three serving tiers behind one endpoint — primary GPU fleet, warm spill, independent fallback — and a fast retryable 503 past absolute capacity.
glossary
loaded
Warm spill
Warm spill is a second serving tier that is already running when the primary fleet fills, so overflow traffic costs latency instead of availability.
capability
1 socket
The synthesis layer for voice agents
The TTS layer of a voice-agent stack: one socket per call, barge-in as one message, 107 ms measured first audio, and lines that price concurrency, not talk.
Your script, this API, thirty minutes — and the recording leaves with you.
Hear it on your own script