Inworld AI
Realtime voice AI platform offering text-to-speech, speech-to-text, a speech-to-speech API and LLM routing.
Inworld AI is a research lab and inference provider that supplies realtime voice AI through APIs. Inworld AI is used by developers to add text-to-speech, speech-to-text and speech-to-speech conversation to consumer applications, and to route chat requests across many language models through one endpoint. Access is by API key, with plans that include monthly usage credits.
Best for Development teams building voice agents, companions, games or learning apps that need low-latency speech models and model routing from a single provider.
- Platforms
- API, Web
- Pricing model
- Credits
- Free plan
- Yes
- API
- Yes
- Made by
- Inworld AI
- Added
Last verified: . Information taken from inworld.ai.
What is Inworld AI?
Inworld AI is a provider of realtime AI models and inference aimed at consumer-facing applications. Its own text-to-speech models, Realtime TTS-2 and TTS-2 Flash, accept bracketed directions inside the text to change tone, pace, volume and pauses, and cover more than 200 languages. Voices can be cloned from 15 seconds of audio or designed from a written description. A Realtime API carries speech in and speech out over one WebSocket or WebRTC connection with turn detection and function calling, while the speech-to-text service adds streaming transcription, voice profiling signals, diarization and word timestamps. The Realtime Router sends LLM requests to models from OpenAI, Anthropic, Google and others at provider rates. The company lists SOC 2 Type II, HIPAA and GDPR among its compliance items and says it has acquired Ultravox.
Key Features
Realtime Text-to-Speech
Synthesises speech with a stated time to first byte under 100 ms for TTS-2 and 25 ms for the Flash variant.
Inline Voice Direction
Reads bracketed instructions placed in the script and adjusts tone, speed, volume, style and pauses accordingly.
Voice Cloning and Design
Builds a custom voice from a 15-second sample, or from a text description of accent, age, tone and energy.
Speech-to-Speech API
Streams full-duplex audio over a single WebSocket or WebRTC link, with adjustable turn taking and tool calls mid-session.
Speech-to-Text with Profiling
Transcribes live or recorded audio and returns emotion, age, accent, pitch and style signals for each chunk.
LLM Router
Routes chat completion requests across more than 100 models with failover and A/B testing and no markup on provider prices.
Screenshots
Use Cases
- Voice Agents
- Power a support or scheduling agent that answers callers with little perceptible delay.
- AI Companions
- Give a companion character a consistent, expressive voice across long conversations.
- Game Characters
- Let players talk with non-player characters that respond in voiced dialogue.
- Language Learning Apps
- Generate spoken practice material in many languages from one cloned or designed voice.
- Model Cost Control
- Send each request to a cheaper or faster language model depending on the user's plan or locale.
Things to know
- Plans are credit based
- Paid tiers bundle a monthly credit amount that can be spent across speech and language model usage, with lower unit rates on higher tiers.
- Compliance options vary by tier
- Zero data retention, HIPAA and BAA are listed as add-ons from the Growth tier, and data residency options appear under Enterprise.
- Focus has moved to voice infrastructure
- The current site presents Inworld as a speech and inference provider and announces the acquisition of Ultravox for speech-to-speech work.
FAQs
What is Inworld AI?
Inworld AI is a research lab and inference provider that offers realtime text-to-speech, speech-to-text, speech-to-speech and LLM routing through APIs.
What is Inworld AI used for?
Inworld AI is used to build voice features into consumer applications such as companions, games, learning tools, wellness apps and customer-facing agents.
Does Inworld AI have an API?
Yes. Inworld AI is delivered through modular APIs with public documentation, an API reference and a playground.
Is Inworld AI free?
Inworld AI has an On-Demand tier that is free to start and includes a limited amount of text-to-speech; usage beyond that and the higher tiers are paid. See the official pricing page for current rates.
How many languages does Inworld AI support?
Inworld AI says its text-to-speech covers more than 200 languages and that one cloned voice can speak across them.
Can Inworld AI clone a voice?
Yes. Inworld AI states that a custom voice can be created from 15 seconds of audio, and voices can also be designed from a text description.
Which language models can Inworld AI route to?
Inworld AI's router reaches models from OpenAI, Anthropic and Google along with more than 100 others through one API.
Who it's for
- Developers
- Voice agent builders
- Game studios
- Companion app teams
- Education platforms
Used in these workflows
- Building an AI-powered app: the tool stack, layer by layerStage 7: Voice, image and video features
Each guide lists the stages of one job and the tools that help at each stage. All guides
Comments
No comments yet. If you have used Inworld AI, tell other readers how it went.
Alternatives to Inworld AI
| Tool | Free plan | Starting price | API |
|---|---|---|---|
 Fish AudioAI voice platform for text to speech and voice cloning with emotion tags, a large voice library and a developer API. |
Yes | — | Yes |
 LazybirdAI voiceover generator that turns scripts into downloadable speech with 200+ voices, voice cloning and an API. |
Yes | — | Yes |
 ListnrAI voice generator for text to speech, voice cloning and narration in a browser-based studio. |
Yes | — | Yes |
 TypecastAI voice generator with emotion-aware text to speech, voice cloning, a mobile app and a developer API. |
Yes | — | Yes |
 Voice.aiAI voice platform combining a real-time voice changer, voice cloning, text to speech and phone voice agents. |
Yes | — | Yes |
Listed in the same categories as Inworld AI. A starting price appears only where we checked it on the maker’s pricing page; a dash means not checked or not published. All Text to Speech tools