Inworld Buys Ultravox: What Changes for Voice-Agent Builders Today, and the Pricing and Voice Promises Nobody Has Made in Writing
Existing Ultravox customers get Inworld's Realtime TTS-2 voices at no extra cost and need no code changes. Deal terms, future pricing, and long-term support for other voice providers were not announced.
TL;DR
Inworld announced on September 30, 2026 that it has acquired Ultravox, a platform developers use to build real-time voice agents. Ultravox's built-in Inworld voices now run on Realtime TTS-2 at no extra cost, with no migration or code changes. For anyone using a voice from another provider, Inworld says only that "nothing changes today". The announcement discloses no deal terms, no pricing commitments, and nothing about the open-weight Ultravox model, so teams building on it should treat pricing and voice choice as open questions.
Inworld, which makes text-to-speech and real-time voice models, has bought Ultravox, a platform for building voice agents that talk in real time. Inworld's announcement is short and mostly about what stays the same. This article separates what Inworld has committed to from what it hasn't.
What Inworld has said
- Ultravox team members have joined Inworld and will keep developing the platform, with ongoing support for existing customers.
- Inworld's Realtime TTS-2 now powers the built-in Inworld voices on Ultravox. The announcement says the upgrade comes at no additional cost, existing voice IDs keep working, and no migration or code change is needed.
- TTS-2 adds natural-language steering, so developers can change pace, tone, and emotional delivery with inline instructions, for example telling a voice to slow down when explaining complicated steps or sound warmer when reassuring.
- Inworld says further speech-to-speech improvements are coming in the following weeks.
- Ultravox co-founder Zach Koch says the two companies partnered on text-to-speech before the deal and that Inworld shares Ultravox's commitment to real-time AI at scale.
Inworld describes Realtime TTS-2 as supporting 200+ languages with sub-100ms latency. Those are Inworld's figures; the announcement does not say how latency was measured.
What the announcement leaves out
| Question | What's known | Source |
|---|---|---|
| Deal price and terms | Not disclosed | Inworld |
| Pricing going forward | No commitment. Beri lists Ultravox's current rates as $0.05 per minute pay-as-you-go and a $100 per month Pro plan, with terms that allow material revisions effective 30 days after notice | Inworld; Beri |
| Voices from other providers | "If you use a voice from another provider, nothing changes today." Nothing is said about later | Inworld |
| The open-weight Ultravox model | Beri says no commitment to keep supporting the MIT-licensed model on GitHub was made | Beri |
| Data handling | Beri flags that it is unclear whether a no-training-on-customer-data clause survives contract changes | Beri |
Beri's reading is that the wording about "today" leaves room to steer customers toward Inworld's own voices over time through defaults rather than removals. That is the author's interpretation, not something either company has said.
What this means if you build voice agents
If you already use Ultravox with Inworld voices, you get a free quality upgrade and nothing to do. If you use ElevenLabs, Cartesia, or another provider's voices through Ultravox, nothing is broken, but ask for the voice-provider and pricing commitments in writing before you sign a longer contract. If your product depends on the open-weight model, pin the version you use now. And if you are choosing a voice-agent platform from scratch, add "who owns it in a year" to your checklist next to latency and price.
Sources
AI Industry Reporter
Priya covers model releases, industry announcements, and the gap between what labs claim and what independent evaluators actually find. She reads the primary source - the paper, the system card, the benchmark org's own statement - before writing a word.
More on AI Voice & Meeting Tools
Fireflies Adds Free, Unlimited Dictation to Every Plan. What Fireflies Talk Does and What the Announcement Doesn't Say
Priya Nair · 4 min
Modulate Raises $25M for Voice AI That Listens to Audio, Not Transcripts. Here's What It Sells and Which Claims You Can Check
Priya Nair · 4 min
Meta Connect 2026: Muse Gets Real-Time Voice, Mac Computer Use, and a Keychain Device
Priya Nair · 5 min