What is Inworld AI?
Inworld AI is a research lab and inference provider for consumer AI applications. Inworld AI builds speech models and serves open language models for developers creating companions, learning products, health and fitness apps, social experiences, personal agents and entertainment. The offering covers both voice interaction and the language-model inference behind conversational applications.
Inworld AI separates speech capabilities into text to speech, speech to text and speech to speech. Hosted inference provides access to models such as Gemma, GLM, MiMo and DeepSeek. Developers can explore the services through a playground, obtain an API key and use the published documentation and API reference.
What Inworld AI can do
Inworld AI provides components for generating speech, understanding spoken input and producing model responses. Realtime TTS-2 supports voice direction in plain language and about 500 languages and dialects. Realtime STT-1 produces streaming transcripts with voice traits, while the speech-to-speech service combines listening, reasoning and spoken replies within a single connection.
- Inworld AI streams generated audio through Realtime TTS-2 and Realtime TTS-2 Flash.
- Inworld AI transcribes speech with emotion, style, age and accent traits.
- Inworld AI supports speech-to-speech sessions over WebSocket or WebRTC.
- Inworld AI lets developers select an LLM for each speech-to-speech session.
- Inworld AI serves open language models through an OpenAI-compatible API.
- Inworld AI provides TTS and STT in LiveKit Agents.
How Inworld AI works
Inworld AI offers an API-key access path alongside an interactive playground. The page includes demonstrations for generating speech, sending language-model prompts, transcribing spoken lines and talking to voice agents. Documentation covers text to speech, speech to text, inference and the API reference, giving developers separate entry points for each service.
Inworld AI supports a combined voice-agent workflow through a speech-to-speech API that listens, reasons and responds over one connection. The service uses the same events and session settings as OpenAI's realtime protocol, which the source says allows existing clients to move with few changes. Developers can also use the voice models independently.
Who Inworld AI is for
Inworld AI targets teams building applications that people use in everyday life, rather than a single standalone chat experience. The page identifies companions, education, health and fitness, social apps, personal agents and entertainment as application areas. Demonstrations also show support interactions, storytelling, language practice and coaching as examples of speech-enabled experiences.
- Inworld AI supports conversational companions and personal agents.
- Inworld AI supports learning experiences such as language tutoring.
- Inworld AI supports health and fitness applications with spoken coaching.
- Inworld AI supports narrated stories and interactive media.
- Inworld AI supports spoken customer-support interactions.
Inworld AI pricing and plans
Inworld AI lists usage-based pricing for speech and inference. On-demand text to speech costs $25 per million characters for Realtime TTS-2 and $15 for Realtime TTS-2 Flash. Growth rates are $12.50 and $7 respectively, while Enterprise rates are $10 and $5. These rates distinguish model choice from plan-level discounts.
Inworld AI prices Realtime STT-1 at $0.10 an hour of audio on every paid plan. Open-model inference starts from $0.07 per million input tokens for Gemma 4 26B. The page also lists speech-to-speech voice pricing at 0.5 ¢ a minute on Enterprise and provides links to pricing, API access and sales contact.
Inworld AI strengths and limitations
Inworld AI combines speech models and hosted language-model inference, with options to use voice components separately or within a connected conversation. A choice of LLM per session and OpenAI-compatible interfaces provide integration flexibility. Regional availability differs by service, however, and the lowest listed voice prices are tied to Enterprise rather than on-demand access.
- Inworld AI provides voice services across the US, Europe, India and Asia Pacific, but LLM serving is in the US.
- Inworld AI reserves the listed lowest text-to-speech rates for Enterprise.
- Inworld AI makes fast mode available on request.
Inworld AI FAQ
What is Inworld AI?
Inworld AI is a research lab and inference provider for consumer AI applications. Inworld AI offers text to speech, speech to text, speech-to-speech conversations and hosted open language models for developers building companions, learning products, personal agents and other consumer experiences.
How much does Inworld AI text to speech cost?
Inworld AI charges $25 per million characters on demand for Realtime TTS-2 and $15 for Realtime TTS-2 Flash. Growth rates are $12.50 and $7, while Enterprise rates are $10 and $5 respectively.
Can Inworld AI voice agents use a different LLM?
Inworld AI lets developers choose the LLM for each speech-to-speech session. The API handles listening, reasoning and spoken replies over one WebSocket or WebRTC connection, using the same events and session settings as OpenAI's realtime protocol.
Which open models does Inworld AI host?
Inworld AI hosts open models including GLM 5.3 Flash, Gemma 4 31B, Gemma 4 26B, MiMo 2.6 Pro and DeepSeek V4.1 Flash. Inworld AI exposes hosted inference through an OpenAI-compatible API.
Which languages and regions does Inworld AI support?
Inworld AI supports about 500 languages and dialects through Realtime TTS-2. Inworld AI provides voice services in the US, Europe, India and Asia Pacific, while LLM inference is served in the US.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




