Ultravox AI

Speech-native voice AI agents. Get started for free at

[Ultravox AI @ultravox_dot_ai]

Agentic use cases are quickly beginning to dominate the world of text-based LLMs, so why are voice AI systems stuck in a world of deterministic flows and node builders?

@zachk on what we need to build truly *agentic* voice experiences: https://t.co/NQ1M0xbOaG

[Kylan @KylanGibbs]

Welcoming @ultravox_dot_ai to the Inworld family today.

Working together more closely over the past year, we found many shared goals and values in our teams.

Bring the platforms together accelerates Inworld's speech-to-speech research and ability to serve reatime AI at scale.

[Inworld AI @inworld]

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.

[Ultravox AI @ultravox_dot_ai]

@inworld Thank you to the developers, customers, and community who have helped shape Ultravox. We’re excited to keep building with you, with more to come in the next few weeks.

[Ultravox AI @ultravox_dot_ai]

We’re excited to announce that Ultravox is now part of @inworld.

We’ve worked closely with Inworld for over a year through our text-to-speech partnership. We share a focus on real-time AI at scale, and that collaboration showed us how much more we could build together.

Read more: https://t.co/vWNNDkatdM

[Ultravox AI @ultravox_dot_ai]

@inworld Key members of our team have joined Inworld to keep developing Ultravox. For customers building with us, your agents keep running, and every built-in Inworld voice on Ultravox is now upgraded to Realtime TTS-2 - same voice IDs, no code changes, and no additional cost.

[Ultravox AI @ultravox_dot_ai]

Get the scoop on 6 trends shaping the voice AI space for the coming year:

1. Emotional AI
2. Proactive Agents
3. Real-Time Multilingual Translation
4. Agentic Automation
5. Multimodal AI
6. On-Device and Edge Architectures

https://t.co/p1p9SrGQa5

[Ultravox AI @ultravox_dot_ai]

👀

[DesignArena @DesignArena]

Introducing Audio Arena

Most existing voice benchmarks are approaching saturation - frontier models are scoring 90%+ on nearly every category.

Today we've open-sourced a suite of 6 static multi-turn benchmarks designed to stress-test speech-to-speech models on realistic customer-facing scenarios.

[Ultravox AI @ultravox_dot_ai]

We're hosting dinner in SF for builders and leaders working in Voice AI. If you're interested in joining us on April 9, please request a seat here:

https://t.co/FG43mMd9Tj

[Ultravox AI @ultravox_dot_ai]

Agentic use cases are dominating the world of text-based LLMs - here's what the voice AI space needs to catch up.

https://t.co/NQ1M0xbOaG

[Ultravox AI @ultravox_dot_ai]

New to voice AI? We created a list of key terms and concepts you're likely to encounter.

Notice something missing? Let us know!

https://t.co/2RXzt6kDCc

[Ultravox AI @ultravox_dot_ai]

Speech is an inherently real-time medium, so for teams seeking to build natural, conversational experiences, latency is table-stakes.

https://t.co/UPVV4nUI5d

[Ultravox AI @ultravox_dot_ai]

3 non-negotiable properties of agentic voice systems:
- Fast
- Fluent
- Fluid

Smarter models alone aren't enough to bridge the gap - voice agents need a harness built for real-time interactions.

https://t.co/NQ1M0xbOaG

[Ultravox AI @ultravox_dot_ai]

Even the most advanced AI models today can’t quite replicate all the nuances of a human-to-human conversation…yet. So what’s still missing?

@kwindla, @bnicholehopkins, and @zachk on evals, model limitations, and how we might measure a truly “natural” voice AI conversation.

https://t.co/E3BN0qRnr1

[Ultravox AI @ultravox_dot_ai]

ICYMI: To celebrate the release of Inworld TTS 1.5, we're giving away up to $100 in promotional credits (that's up to 2,000 call minutes) for Ultravox customers who run production agents with @inworld voices by Feb 28.

Read more: https://t.co/yVHYd4Dt8l

[Ultravox AI @ultravox_dot_ai]

Ultravox v0.7 outperforms *every other major model* on both intelligence and latency for voice AI in the new AIEWF benchmark from @trydaily:

- Ultravox v0.7: 97.7%
- GPT-4.1: 94.9%
- Gemini-2.5-Flash: 93.7%
- GPT-4o: 91.2%
- GPT-Realtime: 86.7%
- Gemini-Live: 86%

The benchmark tests tool calling, knowledge base grounding, and instruction following in a series of multi-turn LLM conversations.

https://t.co/oKx6KwdqM2

[Ultravox AI @ultravox_dot_ai]

In the AIEWF eval, Ultravox's speech-native, real-time model outperformed *all* frontier speech models (including GPT-realtime and Gemini Live), delivering a 97.7% pass rate.

@kwindla, @bnicholehopkins, and @zachk discuss how Ultravox closed the performance gap between speech and text models.

https://t.co/7hQMB6Ppud

[Ultravox AI @ultravox_dot_ai]

Speech-to-speech models are poised to displace the voice AI component stack — and benchmarks like the new AIEWF eval from @kwindla prove it.

https://t.co/4QIha0YfZm