Voice agents
Useful when you already have the orchestration working and the TTS bill is the next thing making the economics ugly.
OpenClaw guide
Use SayWell with OpenClaw in one config block.
If you already use OpenClaw's OpenAI TTS provider path, the migration story should stay boring: keep the provider, swap the base URL, test the voice, and compare the cost math.
OpenAI-compatible$5 / 1M charsEarly access
Game dialogue
Useful when you need cheap iteration for NPCs, branching dialogue, or cinematic placeholders before final polish.
Cost replacement
Useful when you do not want to rework your app architecture just to try a cheaper TTS backend.
OpenClaw setup
Keep the provider path. Change the endpoint.
This is the whole point of the compatibility story. You do not need a bespoke SayWell client to see whether the price and voice quality are interesting.
{
messages: {
tts: {
auto: "always",
provider: "openai",
summaryModel: "openai/gpt-4.1-mini",
modelOverrides: {
enabled: true,
},
providers: {
openai: {
apiKey: "saywell_api_key",
baseUrl: "https://saywell.io/v1",
model: "studio",
voice: "aria",
},
},
},
},
}Setup flow
How this guide should work in practice.
The page exists to make the integration legible enough that a builder can try it, judge the economics, and decide whether early access is worth it.
01
Request a SayWell alpha key
Join the early-access list or email directly for a manual alpha key while the product surface stays intentionally small.
02
Keep the OpenAI provider path
Use OpenClaw's built-in OpenAI speech provider instead of inventing a separate integration surface.
03
Point baseUrl at SayWell
The only operational change is the endpoint. That keeps the migration story simple and makes cost comparisons legible.
04
Choose your model and voice
Pick the SayWell voice you want and test one outbound reply before wiring it into a larger automation or call flow.
What stays the same
OpenClaw stays on `provider: "openai"`
Your TTS orchestration remains in one config surface
Model and voice selection stay in the same provider block
What changes
Use your SayWell key in `providers.openai.apiKey`
Point `providers.openai.baseUrl` to `https://saywell.io/v1`
Pick the SayWell model and voice you actually want to test
Notes
If you already keep a second provider around for fallback, do not throw that away. The goal here is to validate whether SayWell deserves the primary slot in your stack, not to force a more dramatic migration than necessary.
Current alpha support: `model=studio`, voices `aria`, `sol`, and `vale`, plus `mp3`, `wav`, and `opus` output from a real live endpoint at `https://saywell.io/v1`.
Quick pricing math
Judge the economics before you judge the roadmap.
This section is intentionally simple. If the budget math is not compelling enough to make you curious, there is no point pretending the integration story will save it.
Free tier
100K chars / month
1M characters
$5
5M characters
$25
10M characters
$50
Warm onboarding guide
Aria
Balanced and clear for product tours, support flows, and tutorials.
“Welcome to SayWell. Type your script once, choose a voice, and ship production-ready audio without chasing credits.”
Confident operator
Sol
Sharper cadence for voice agents, assistants, and menu systems.
“Your API key includes one hundred thousand free characters every month, with commercial use included from day one.”
Cinematic narrator
Vale
A broader performance range for trailers, games, and promo reads.
“Build the scene, stream the dialogue, and keep every line item predictable from prototype to launch.”
FAQ
The questions this guide should answer quickly.
If any of these still feel fuzzy, the page is doing a bad job and we should tighten it before spending money on traffic.
Do I need to rewrite my OpenClaw setup?
No. The point of the guide is to keep the migration narrow: stay on the OpenAI TTS provider path, swap the API key and base URL, then test one voice.
Can I still keep another provider as fallback?
Yes. If you already rely on ElevenLabs or another provider for fallback, keep that block in your OpenClaw config and use SayWell as the primary OpenAI-compatible path.
Is SayWell fully open right now?
No. This guide is for early-access validation and onboarding. The waitlist lets us see which builders actually want the integration before we widen access.
What is live in the alpha API today?
The live alpha supports `POST /v1/audio/speech`, the `studio` model, `aria` / `sol` / `vale` voices, and `mp3` / `wav` / `opus` output. The goal is to make OpenClaw testing real before broadening the surface area.
Outreach asset
This page is the thing we send before asking for a docs PR.
It gives us a neutral reference asset for community posts, maintainer outreach, and any future docs contribution without forcing OpenClaw to host vendor marketing on day one.
OpenClaw builders
Request alpha access for your OpenClaw stack.
If this integration story is interesting, join the list. We are onboarding the first wave manually so we can watch real usage, tighten the compatibility surface, and decide what to widen next.
OpenAI-compatible endpoint
Manual alpha API keys
Priority migration updates