Inworld AI
Realtime TTS, STT, and LLM routing for interactive characters and voice apps
About Inworld AI
Inworld AI started as a character engine for games and now sells realtime voice infrastructure: Realtime TTS-2 and TTS-2 Flash text-to-speech, Realtime STT with voice signals such as emotion and pitch, a speech-to-speech Realtime API, and an LLM router across 100+ models. TTS supports 200+ languages, natural-language steering tags, instant cloning and voice design, and streams over WebSocket. Plans are credit-based, from free on-demand use to $1,500/month Growth and custom enterprise deals.
Pros & Cons
Pros
- Per-character rates well below most premium TTS vendors, with volume discounts that scale to $5/1M characters
- Concurrency limits are generous for the price (50 concurrent requests on the $100 Builder plan)
- Inline steering and non-verbal tags give direct control over delivery without SSML
- One platform covers TTS, emotion-aware STT, speech-to-speech, and LLM routing for character pipelines
- Long history in game characters, with a dedicated games and media offering
Cons
- Latency figures are server-side P99 and exclude network round trip, so real end-to-end numbers are higher
- Engine integration is through standard APIs; you build the Unity or Unreal glue yourself
- Professional voice cloning is an add-on until the Growth plan
Key Features
Realtime TTS-2
Flagship expressive TTS with under 100 ms P99 time to first byte (server-side, excluding network), per the vendor.
Realtime TTS-2 Flash
Lower-cost, lower-latency model with a vendor-stated 25 ms P99 time to first byte.
Natural-language steering
Bracketed instructions such as [Speak sadly] and non-verbal cues such as [laugh], [sigh], and [breathe] inline in the text.
Voice cloning and voice design
Clone a voice from 5 to 15 seconds of audio or design one from a text prompt, then localize it across 200+ languages.
Realtime STT with voice signals
Speech-to-text that returns emotion, age, accent, pitch, and vocal style signals with confidence scores.
Realtime API
Speech-to-speech sessions with semantic turn detection and barge-in for interruptible dialogue.
LLM Router
Route each request across 100+ LLMs by metadata such as scene, platform, or locale, with automatic fallback.
Engine-agnostic APIs
REST, WebSocket, and WebRTC APIs that can be called from Unreal, Unity, or custom engines.
Pricing
- Up to 70 min TTS included
- TTS-2 $25/1M chars, Flash $15/1M chars
- 100 custom voices
- 5 concurrent requests
- Commercial license
- $25 credits per month
- TTS-2 $20/1M, Flash $10/1M
- 500 custom voices
- 10 concurrent requests
- $100 credits per month
- TTS-2 $17.50/1M, Flash $9/1M
- 3,000 custom voices
- 50 concurrent requests
- $300 credits per month
- TTS-2 $15/1M, Flash $8/1M
- 10,000 custom voices
- 150 concurrent requests
- Priority email support
- $1,500 credits per month
- TTS-2 $12.50/1M, Flash $7/1M
- 30,000 custom voices
- 500 concurrent requests
- 1 professional voice clone
- TTS-2 as low as $5/1M chars
- Custom limits
- SLA and DPA
- EU and India data residency
Best For
Game NPC dialogue
Voice LLM-generated NPC lines in real time, routing cheap barks and story-critical scenes to different models.
AI companion apps
Expressive, low-cost voice for consumer companion products with many simultaneous users.
Voice agents
Speech-to-speech agents with turn detection and interruption handling.
Similar Tools
ElevenLabs
AI voice generator and voice agents platform
Murf AI
AI voice generator with 200+ realistic text-to-speech voices
TTSOpenAI
Advanced AI voice engine for natural text-to-speech
LOVO AI
AI voice generator and video editor with 500+ voices in 100+ languages
Groq
Ultra-fast AI inference powered by custom LPU silicon
Fliki
Turn text into videos with AI voices in minutes