September 30, 2026
Ultravox agent voices now take text directions: Inworld integrates Realtime TTS-2
Built-in Inworld voices now support delivery controls through text instructions, with no migration, code changes or extra charge.
🚨 AI News | TestingCatalog
@testingcatalog
Inworld has acquired Ultravox, a platform for building voice agents that listen, reason, call tools and respond in live conversations. The platform also manages turn-taking and handles interruptions. Together, the companies aim to integrate speech understanding, reasoning and expressive voice synthesis more closely in agents that take speech as input and respond with speech. > Inworld's built-in voices in Ultravox now run on Realtime TTS-2. > Existing voice IDs continue to work without migration or code changes. > TTS-2 also lets developers control how Ultravox agents speak using plain-language instructions.
We’re excited to announce that @ultravox_dot_ai is now part of Inworld. Ultravox is the platform developers use to build real-time voice agents. We've worked with the team for a while through our TTS partnership, and today members of the team that built it are joining Inworld to keep developing it.
· 4.8K views
On Sep 30, 2026, Inworld announced its acquisition of Ultravox. Employees joining Inworld will continue developing the voice AI agent platform.
Ultravox supports real-time conversations: an agent listens to the other person, reasons, calls tools and responds with speech. The platform manages turn-taking and handles interruptions.
Voices follow instructions. Since Jan 21, Ultravox has offered an Inworld TTS 1.5 integration with preset voices in 15 languages and support for users' cloned voices. Built-in Inworld voices now run on Realtime TTS-2, while existing voice IDs remain valid.
With TTS-2, you can specify pace, tone and emotional delivery directly in the utterance text. Instructions in square brackets, such as `[Speak softly]` or `[Speak sadly]`, change delivery even within a single utterance. Using them is optional.
Setup in the interface. In app.ultravox.ai, open Voices, filter by the Inworld provider and listen to samples. Then assign a voice to an existing or new agent in Agents. Built-in voices do not require a separate TTS provider API key; those keys are required for ExternalVoices.
You can retrieve the voice list programmatically with a `GET https://api.ultravox.ai/api/voices` request using the `X-API-Key` header. The `provider` parameter filters the list by the value `inworld`.
As of Sep 30, the Pay As You Go plan costs $0 per month and includes the first 30 minutes of calls for free. After that, calls cost $0.05 per minute, with up to 5 concurrent calls available. The number of calls in Playground is unlimited.
Inworld and Ultravox plan to bring speech understanding, reasoning and expressive speech synthesis closer together in voice agents.
