Integration · Vapi

Saidwire speech-to-text for Vapi

Q: Do I need to host anything?

No. Saidwire runs the custom-transcriber server at wss://vapi.saidwire.com/vapi/transcriber; you only change the assistant's transcriber setting.

Make Wire-1, Saidwire's real-time English speech-to-text, the transcriber of any Vapi assistant with one change to the assistant's configuration. Your prompt, model, voice and phone numbers stay as they are.

Vapi supports custom transcribers: for each call it streams the call audio to a WebSocket server and reads the transcripts back. Saidwire runs that server for you at wss://vapi.saidwire.com/vapi/transcriber. It forwards the caller's audio to the Saidwire API and returns partial transcripts while the caller speaks, then a final transcript when they stop. Only the caller's side of the call is transcribed and billed, not your assistant's.

On phone-codec audio in Coval's STT benchmark, Wire-1's mean time to final segment was 38 ms (30-day window to Nov 9, 2026). Source: Coval model page, 30-day window to Nov 9, 2026, 2,857 samples, US-hosted endpoint. Add your own network round trip. Method and how to reproduce.

1. Add your Saidwire key to Vapi

In the Vapi dashboard, open Integrations → Server Configuration → Add Custom Credential:

Field	Value
Authentication Type	Bearer Token
Credential Name	`Saidwire`
Token	your Saidwire key (`sk_live_...`) from saidwire.com/dev
Header Name	`Authorization`
Include Bearer Prefix	on

Save, then copy the credential's ID.

2. Set the assistant's transcriber

In the assistant's settings, or through the API, set:

{
  "transcriber": {
    "provider": "custom-transcriber",
    "server": {
      "url": "wss://vapi.saidwire.com/vapi/transcriber",
      "credentialId": "YOUR_CREDENTIAL_ID"
    }
  }
}

The same change with the Vapi API:

curl -X PATCH "https://api.vapi.ai/assistant/$ASSISTANT_ID" \
  -H "Authorization: Bearer $VAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"transcriber":{"provider":"custom-transcriber","server":{
        "url":"wss://vapi.saidwire.com/vapi/transcriber",
        "credentialId":"YOUR_CREDENTIAL_ID"}}}'

3. Test it

Start a web call or a phone call with the assistant and speak. The transcripts appear in the call log.

Good to know

Saved vs. transient assistants. Vapi sends a stored credential only with saved configuration. If you create the assistant inline in an API call, put the key in the server headers instead: "server": {"url": "wss://vapi.saidwire.com/vapi/transcriber", "headers": {"Authorization": "Bearer sk_live_..."}}.
Turn-taking. Saidwire sends partial transcripts while the caller speaks and finalizes after a short pause (about 300 ms of caller quiet). Vapi's own start-speaking settings then decide when the assistant replies.
Fallback. If Saidwire can't take a call (for example, the key's hours are used up), the connection closes with an error, and a Vapi fallbackPlan on the transcriber can switch that call to another transcriber.
Concurrency. Each call uses one Saidwire stream. Trial keys allow 4 at once; write to support for more.
Privacy. The connector doesn't store audio, transcripts or keys.

Troubleshooting

What you see	Fix
No transcripts, and the call log shows a 401	The credential isn't attached. Check that it's a Bearer Token credential and that the assistant references its ID. For transient assistants, use `headers`
A `402` error	The key's hours are used up. Add a card at saidwire.com/dev
A `429` error	More calls at once than the key allows. Ask support for a higher limit
The assistant answers before the caller finishes	Raise the wait time in Vapi's start-speaking settings

Questions

Do I need to host anything?

No. Saidwire runs the custom-transcriber server at wss://vapi.saidwire.com/vapi/transcriber; you only change the assistant's transcriber setting.

Is my assistant's own voice billed?

No. Only the caller's side of the call is transcribed and billed.

What does it cost?

120 free hours with no card and no deadline, up to 4 streams at once, then $0.25 per audio hour, billed per second of caller audio.

Which languages?

English only. Use another transcriber for assistants that need other languages.

Questions or stuck? Email support@saidwire.com with the request ID from the failed call.

Pipecat, LiveKit, Vapi, OpenAI and the other names on this page belong to their respective owners. They are named only to describe compatibility; no partnership or endorsement is implied. Coval benchmark figures are from Coval's public model page; Coval does not endorse Saidwire.