The Pulse
Inworld Buys Ultravox to Add Voice Agents to Its Speech Stack
Inworld acquired Ultravox, bringing its real-time voice-agent platform together with Inworld’s speech models and inference infrastructure. Built-in Inworld voices on Ultravox now use Realtime TTS-2, with existing voice IDs retained and no a
AI.info Team · Written with AI ·
“I'm incredibly proud of what the Ultravox team built: real-time voice AI that customers depend on every day. When we partnered with Inworld on text-to-speech, we saw a team that moves fast and shares our commitment to delivering real-time AI at scale. I believe Inworld is the right home for Ultravox and the right team to take it further.”
Zach Koch, co-founder and former CEO, Ultravox
Inworld announced the acquisition of Ultravox on September 30, bringing a developer platform for real-time voice agents into the same company as its speech models and inference infrastructure. Ultravox team members who joined Inworld will continue developing the platform, the company said. It did not specify how many joined.
Ultravox adds the agent layer
The companies make different parts of a voice interaction. Ultravox provides speech understanding, reasoning and task completion, as well as tools for handling turn-taking and interruptions in live conversations. Inworld supplies speech models, model serving and real-time inference, including Realtime TTS-2 and Realtime STT-1.
Combining those capabilities puts the agent platform alongside the technologies that process incoming speech and generate spoken replies. Inworld says the goal is to improve the full exchange: understand what a user says, help complete a task and respond in a voice suited to the moment. The acquisition therefore brings Inworld closer to owning both the speech components and the developer-facing product that connects them.
Built-in Inworld voices move to TTS-2
The first product change is already live on Ultravox. Built-in Inworld voices now run on Realtime TTS-2, and customers using those voices keep their existing voice IDs without changing code. Inworld says the upgrade adds no charge.
TTS-2 adds optional natural-language controls for a voice’s pace, tone and emotional delivery. Developers can, for example, direct a voice to slow down during a complicated explanation or sound warmer when reassuring a user. Inworld says the speaker identity stays the same, though pacing and intonation may differ with the updated model.
Inworld describes TTS-2 as a top-ranked voice model on Artificial Analysis.
Other voice providers remain in place
Ultravox customers can keep using their accounts, and Inworld says existing agents require no migration as part of the acquisition. Customers using voices from other providers see no change for now. Support will continue through Inworld and members of the Ultravox team who joined the company.
That distinction matters for a platform whose users may have built agents around different voice providers. Inworld’s own voices get the immediate TTS-2 upgrade, while the announcement says third-party voices remain unaffected. It describes a product change, not a requirement to replace other providers.
The next test is broader than a voice upgrade
The acquisition gives Inworld a direct route from speech technology into a product developers already use to build and run voice agents. For Ultravox customers, the stated near-term benefit is specific: an upgraded Inworld voice option without extra cost or code changes. Inworld says further capabilities and improvements for Ultravox are planned, but the announcement does not detail them.
The first integration connects Inworld’s voice generation to Ultravox’s agent platform; the larger pitch is to bring speech understanding, reasoning and expressive speech closer together. Whether that combination makes development or live conversations better will depend on changes users can test beyond the TTS-2 upgrade.