Inworld AI

AI built for people.

[Inworld AI @inworld]

Inworld is your new favorite inference provider.

We already process tens of trillions of tokens a month because we understand consumer AI different, and we've built our inference from the ground up for it.

Today, the inference top consumer AI apps run on is opening up.

[Inworld AI @inworld]

Inworld’s international presence keeps growing. We’re hiring across Europe, with roles in our Science and ML Engineering teams in Germany, Switzerland, Serbia, and the UK.

We’re looking for research scientists and ML engineers to help build our speech models and the inference systems behind realtime AI applications.

If that sounds like your kind of work, we’d love to hear from you.

Explore the roles and come work with us: https://t.co/LLb0t75qH7

[Inworld AI @inworld]

Our @livekit partner story explores how Inworld’s speech models and LiveKit’s realtime infrastructure help developers build voice applications that can scale with their users.

Through Inworld’s integration with the LiveKit Agents framework, developers can use pre-built or custom voices alongside their choice of LLM and speech recognition models, with LiveKit handling realtime communication across web and mobile apps.

The accompanying video takes a closer look at expressive delivery, including how TTS-2’s voice steering helped the team bring Expressive Mode to LiveKit Agents.

“Inworld's TTS-2 marks a real step forward in emotionally expressive voice synthesis. When combined with the conversational intelligence of LiveKit agents, it enables interactions that feel genuinely human, responsive, nuanced, and alive in ways that feel natural.” - David Zhao, CTO, LiveKit

Watch the video or read the full partner story for more on our work together: https://t.co/4MTeupx93h

[Inworld AI @inworld]

@cekuraAi's new Text-to-Speech Bench tested 26 models across 30,564 requests.

To keep the comparison consistent, every model read the same 104 sentences three times from the same cloud region, with automated speech-to-text scorers estimating reading accuracy.

Inworld stood out for both response speed and reading accuracy. A few results from the evaluation:

- TTS-2 Flash started speaking in a median of 46ms when given a full sentence.
- TTS-2 scored 93.4% on estimated reading accuracy, with Flash at 93.2%, covering details such as phone numbers, dates, codes, and medical terms.

Thanks to the Cekura team for making the methodology and audio samples available so everyone can explore the results and hear the differences themselves.

Take a look at the benchmark and have a listen: https://t.co/8YgjpTrzfT

[Inworld AI @inworld]

Latitude wanted every story beat in Voyage to be narrated, with a distinct, consistent voice for each character. Its existing voice provider made that too expensive, while its separate inference service slowed down as traffic grew.

We partnered with the team to make always-on narration affordable and serve the models behind Voyage’s game mechanics and stories.

Key results:
- Voice at roughly a quarter of the cost of their previous provider
- TTS 3–5x faster than another provider in Latitude’s production tests
- More than 15 million TTS calls in six weeks
- More than 25 million inference calls over the same period

With Inworld’s support, Latitude made narration a default part of Voyage. The team also runs its own fine-tuned story models on Inworld without having to manage the infrastructure to serve them.

Read the full case study: https://t.co/iwZjK6IRGF

[Inworld AI @inworld]

Status, the AI social simulation game built by Wishroll, was growing fast. Its five-person team needed to bring AI costs down without compromising the experience players expected.

We partnered with them to make that growth sustainable.

Key results:
- 95% reduction in AI costs
- Scaled to 500,000+ daily active users
- 35x revenue growth to tens of millions in annual revenue
- Sub-second response times, even at peak demand

With Inworld’s support, the team could keep expanding the game’s audience while spending more engineering time on new features.

“Working with Inworld has been the most positive experience I've had with another company... They've been incredibly hands-on, responsive, and fast. It's like having an external ML team dedicated to helping us solve our problems.”
- Fai Nur, CEO and co-founder of Wishroll / Status

Read the full case study for a closer look at how we worked together.: https://t.co/ompFx193O1

[fai nur @faionur]

@thestatusai has scaled to tens of millions in annualized revenue. Working with @inworld, we cut our AI costs by about 95% while maintaining or improving content quality.

When you build consumer AI, you want people to be obsessed with your app. Then you have to afford that obsession.

Our users spend an average of over an hour and a half a day on Status. They’re building relationships with characters, creating stories, and getting absorbed in these worlds. Every interaction needs to feel worth coming back for.

Inworld’s team got into the details with us: which models to use, how to structure prompts, and how to test whether the content was actually funny and engaging. They’ve been hands-on and fast, and we’ve kept working together to improve performance and lower costs.

That gave our team more room to keep shipping. More features, more things for players to do, and a business that could support the experience we wanted to build.

There’s so much opportunity in consumer AI when you understand what people enjoy. Making the economics work gives you room to keep finding more of it.

Read more on how we scaled with Inworld: https://t.co/EtWBdjVNz3

[Inworld AI @inworld]

@thestatusai @LatitudeGamesAI Choose a service tier per request, add model fallbacks, or bring your own weights for a dedicated deployment.

Start with our API, or try a model in the browser with no API key.

AI built for people runs on Inworld. Who are you building for?

[Inworld AI @inworld]

Inworld is your new favorite inference provider.

We already process tens of trillions of tokens a month because we understand consumer AI different, and we've built our inference from the ground up for it.

Today, the inference top consumer AI apps run on is opening up.

[Inworld AI @inworld]

With @thestatusai, we worked on prompts, evaluations and serving to cut AI costs by over 95% while maintaining or improving content quality.

@latitudegamesai ran 25M+ inference calls in its first six weeks on Inworld.

[Inworld AI @inworld]

We recently announced our acquisition of Ultravox. GeekWire covered the news, with some background on the company and its journey from Fixie to Ultravox.

If you missed the announcement, read the story here: https://t.co/IWShmrKzuK

[Inworld AI @inworld]

We recently shared our perspective in USA TODAY on the infrastructure behind consumer AI.

The article explores what it takes to keep AI applications fast and affordable as people use them more often.

It covers how serving costs change as usage grows and why realtime voice requires coordination across multiple systems. There’s also a look at Inworld’s origins in interactive experiences and how that work shaped our focus on realtime AI for consumer-facing applications.

Read the full story: https://t.co/KPKYiZteLB

[Inworld AI @inworld]

In his blog post “WTF is Voice Steering?”, our Head of Research, Aleksey Tikhonov, explains how voice steering in TTS-2 lets you give a speech model instructions about how to deliver a line.

The piece focuses on how we measure whether those instructions actually work. Aleksey explains where automated audio judges fell short in our tests and how the team measures changes in the speech itself.

He also explores why instruction wording matters and why the same direction can produce different results across voices. These measurements help the team track changes between model versions, with limits on what they can tell us.

Read the full piece: https://t.co/I1Bzzqn4kj

[Inworld AI @inworld]

Muse Spark 1.3 is live on Inworld Realtime Router.

- Meta’s reasoning and coding model, closed weights
- 1M token context window
- $1.25 in / $4.25 out per 1M tokens
- ~25% fewer tokens than Spark 1.2 in Meta’s internal coding tests

[Voyage @playVoyage]

In Voyage, the narrator and every character you meet have their own voice, generated live for whatever you do.

Our SOTA World Engine gives every line of dialogue a delivery direction (warmly, breathless, urgently) and @inworld TTS performs it live.

Here's what that unlocked:
4x faster voice at a quarter of the cost. Those savings helped us offer unlimited play to all our subscribers.

Our founder and CEO @nickwalton00 sat down with the Inworld team to talk about why a robotic voice breaks a story, and how we made every voice in Voyage feel real.

[Inworld AI @inworld]

Join us in Berlin on Friday, October 23 for Backchannel Berlin, an afternoon of conversation about the research on speech-to-speech models.

Our team will share how we're building and serving full-duplex speech-to-speech in realtime, including the things that haven't worked. Then the floor opens: bring your own work or a problem you're stuck on, or just come and listen.

We'll be getting into turn-taking and interruptions, cascaded versus end-to-end, keeping an LLM's reasoning when you teach it to speak, cross-lingual transfer, and what we should actually be measuring for conversations.

If you work on speech-to-speech, TTS, ASR, audio codecs, conversational modelling, LLM serving or evaluation, we'd love to have you. PhD students welcome.

Food and drinks throughout. Register at https://t.co/jpmHj1ngQ7