Regions

Overview

The Speechify API answers on three hosts. They take the same API key, accept the same requests and return the same responses. They differ only in which region runs speech synthesis.

api.sws.speechify.com is a legacy host, not a region: it stops serving on 19 January 2027 (see the changelog), so move any integration on it to api.speechify.ai.

What the host decides

The host chooses the region for the text-to-speech and voice endpoints only.

Pin a region

Set the regional host as your client’s base URL. Nothing else in your integration changes.

import os
from speechify import Speechify
client = Speechify(
    token=os.environ["SPEECHIFY_API_KEY"],
    base_url="https://eu.api.speechify.ai",
)
audio = client.audio.speech(
    input="Hello from Europe.",
    voice_id="geffen_32",
    model="simba-3.2",
    audio_format="mp3",
)

The WebSocket input stream takes the host the same way: wss://eu.api.speechify.ai/v1/audio/stream/ws. Integrations that accept a base URL take the regional host in place of the default, such as base_url in Pipecat and baseURL in the Vercel AI SDK provider.

How api.speechify.ai routes

Today api.speechify.ai serves every request from US East, wherever it comes from. It is the host that will route by location: when it starts sending requests to the region nearest the caller, the changelog will say so. If your integration depends on a fixed region, use a regional host rather than the default.

When a model has no capacity in the region

A request is never refused because its model has no capacity in the host’s region. The synthesis service runs in that region and sends the model work to the nearest pool that serves your model:

1. A pool in the same region, when the model has one.
2. Otherwise a pool in another region, usually US East.

Which models have capacity in Europe changes as we add it. If a pool fails before any audio has been sent, the request moves to the next pool. Once audio has started, the request stays where it is.

A pinned region is therefore not faster on its own: when the model runs elsewhere, the request still crosses to it. Measure first audio from where your code runs, on each host, before you choose; Latency shows how.

Availability

A regional host serves only from its own region, and never hands a request to another region’s service. If that region has an outage, requests to its host fail until it recovers. To keep serving through one, send the retry to another host. Incidents are posted on status.speechify.ai.

Limits and billing

What a region does not change

Regional hosts choose where synthesis runs. They do not keep your data in a region, and they do not make the API a data-residency service. On every host:

• Your account, API keys, workspaces, projects and voices are stored in the United States.
• Every request is authenticated against account data held in the United States.
• Usage records, billing and request logs are kept in the United States.
• The text you send can be processed in the United States: for content checks, and whenever the model runs there.

If you need your data stored and processed only within the EU, the regional hosts do not provide that.

Related