Inworld AI

AI voice / speech API platform (TTS, STT, Realtime)

Inworld AI provides real-time voice AI infrastructure — text-to-speech, speech-to-text, LLM routing, and speech-to-speech Realtime API — for games, apps, and interactive experiences. Originally known for NPC character engines, Inworld pivoted in 2025-2026 to a developer-first voice AI platform with sub-200ms latency, 220+ LLM models via Router, and tiered API pricing scaling to enterprise.

Alternatives & similar tools

Related picks from editorial notes.

Also mentioned

ElevenLabs
PlayHT
Google Cloud Text-to-Speech

Features & details

Full overview from our catalog (read-only reference).

Category (short)

AI voice / speech API platform (TTS, STT, Realtime)

Sponsored note

Target audience

Game developers adding real-time character voice dialogue, App builders needing low-latency TTS/STT APIs, Teams building voice-enabled AI agents and companions, Not ideal for: non-technical users wanting a consumer app, video avatar generation, or simple one-off voiceovers

Best for

Pricing notes

On-Demand: free — up to ~70 min TTS, 100 custom voices, community support (inworld.ai/pricing, verified July 2026). Creator: $25/mo — $25 credits/mo, TTS-2 at $20/1M chars. Builder: $100/mo — $100 credits, TTS-2 at $17.50/1M chars, 3,000 voices. Developer: $300/mo — $300 credits, TTS-2 at $15/1M chars, 10K voices. Growth: $1,500/mo — $1,500 credits, TTS-2 at $12.50/1M chars. Enterprise: custom, TTS-2 as low as $5/1M chars. Credits roll over 3 months on paid plans. LLMs billed at provider cost via Router.

Pros & cons

Editorial notes to help compare fit before opening the vendor site.

Pros

Aggressive TTS pricing significantly below ElevenLabs at scale
Sub-200ms Realtime API latency for interactive applications
LLM Router with 220+ models at zero markup
Generous free tier for prototyping (~70 min TTS)
Credit rollover up to 3 months on paid subscriptions

Cons

Pivot from NPC character engine confused existing user base
Complex tier structure requires cost estimation before committing
Voice AI focus may not suit teams expecting visual avatar products
Professional voice cloning is paid add-on, not included
Enterprise features (HIPAA, on-prem) require Growth tier minimum

Review notes

Context from the listing review and editorial research.

Inworld's 2026 pivot to voice AI APIs reflects market reality — TTS/STT infrastructure pays better than consumer NPC demos. New pricing undercuts ElevenLabs significantly at scale ($5 vs ~$100/1M chars enterprise). Developers praise latency and Router breadth; game studios miss the original character engine focus. Free tier generous for prototyping. Production apps land on Developer ($300) or Growth ($1,500) tiers.

Extended features

In-depth description and capability notes.

Realtime TTS-2 — High-fidelity text-to-speech from $5–$25 per 1M characters by tier Speech-to-Text — Audio transcription from $0.10/hr at scale Realtime API — Combined speech-to-speech sessions with sub-200ms latency LLM Router — Access 220+ models at provider cost with no markup Voice cloning and design — Custom voices from 100 (free) to 30,000 (Growth tier) Workspace collaboration — Team sharing and management on Builder+ Compliance add-ons — HIPAA, BAA, zero data retention on Growth/Enterprise

Use cases

Inworld fits developers building voice-enabled apps, games, AI agents, and interactive experiences needing low-latency TTS/STT at scale. Game NPC voice dialogue and character speech AI companion and chatbot voice interfaces IVR and customer service voice agents Interactive education and tutoring applications Real-time speech-to-speech conversational AI products