Inworld TTS Review, Pricing & Alternatives (2026)

Inworld TTS is Inworld AI’s hosted speech service, available through its Portal, playground, SDKs, and HTTP APIs. It targets developers building games, companions, assistants, and interactive applications rather than consumers seeking a desktop narration tool. Inworld launched its first TTS models in 2025, including TTS-1 and the larger TTS-1-Max. The technical report describes 1.6 billion parameter TTS-1 and 8.8 billion parameter TTS-1-Max systems. Current product documentation centers on Realtime TTS-2 and TTS-2 Flash. Inworld lists API access, commercial licensing, GDPR, and SOC 2 Type II coverage across its plans. Enterprise customers can add custom limits, data residency, an SLA, and a DPA. A typical workflow uses the TTS Playground for voice selection or design, followed by API-key authentication and synthesis through HTTP, streaming, or WebSocket requests.

A typical workflow begins in the TTS Playground, where developers compare stock voices, create voice designs, or preview instant clones. Production code creates an API key and sends text, a voice, and a model to the synthesis service. Non-streaming HTTP returns complete audio, while HTTP streaming and WebSocket stream chunks during generation. WebSocket is the documented choice for the lowest latency. The JavaScript SDK is installed as @inworld/tts, and the Python package is inworld-tts. Output formats include MP3, Opus, WAV, PCM, and FLAC through compatible endpoints. Async jobs handle long text, with limits up to 100,000 characters, while Batch costs 20 percent less per character. Realtime TTS-2 supports natural-language steering, inline non-verbal cues, timestamps, cross-lingual voice use, and synthesis across 15 languages.

Realtime TTS-2 is Inworld’s expressive model, with natural-language steering, inline non-verbal cues, cross-lingual voice use, timestamps, and production-quality synthesis across 15 languages. Current product material lists 200-plus languages and locales for the TTS-2 family. Realtime TTS-2 Flash reports 20 ms server-side P90 time to first audio versus 100 ms for TTS-2. Flash supports the same broad language coverage and non-verbal tags, while steering instructions and professional voice cloning are supported on TTS-2 only. Earlier releases included TTS-1 and TTS-1.5. The technical report describes Transformer-based autoregressive models, 48 kHz output, short-reference voice conditioning, and audio markups. Production access includes HTTP and WebSocket streaming, asynchronous jobs for long text, batch processing, and SDKs for JavaScript and Python.

On-Demand costs $0 monthly plus usage, with TTS-2 at $25 per million characters and Flash at $15. It includes up to 70 minutes of TTS, 100 custom voices, two professional voice clones, voice cloning, voice design, API access, and a commercial license. Creator costs $25 monthly and includes $25 in credits, with rates of $20 and $10. Builder costs $100, Developer $300, and Growth $1,500, with lower rates and higher limits. Concurrency increases from five requests on On-Demand to 500 on Growth. The playground accepts 2,000 characters on On-Demand and 40,000 on paid plans. Enterprise pricing is custom, with TTS-2 as low as $5 per million characters and Flash below $5. Paid plans increase concurrent request limits from 10 on Creator to 50 on Builder, 150 on Developer, and 500 on Growth.

Inworld’s main constraints are cloud dependence, character metering, plan complexity, and feature separation between TTS-2 and Flash. ElevenLabs is a close alternative for creator-oriented voice discovery and cloning, while Inworld emphasizes realtime API integration and application workflows. Cartesia competes on realtime speech, whereas Azure AI Speech and Google Cloud Text-to-Speech offer wider enterprise cloud-service portfolios. Kokoro is a different choice for teams willing to self-host an open model instead of paying a managed per-character service. Inworld provides HTTP and WebSocket streaming, asynchronous jobs, Batch processing, SDKs, and an OpenAI-compatible endpoint for integrating speech into applications. Pick Inworld if your product needs managed realtime speech, API integration, and expressive controls.