Helps consumer AI teams build realtime voice apps at lower cost
Commission may apply at no extra cost
Fast, low-cost voice for realtime apps
Inworld AI puts low-latency speech-to-text and text-to-speech behind one API, with no markup on model routing and voice rates below the providers it compares against. The trade-off is a developer-only, fully metered setup with no free tier or refund window. Overall, a strong pick for teams that build in code and prize voice latency and cost.
Inworld AI is a realtime voice and model-routing platform for consumer-facing AI apps. It combines speech-to-text, text-to-speech, and one speech-to-speech API over WebSocket or WebRTC. A router sends requests to models from major providers with no markup added.
It suits product and engineering teams building voice agents, companions, and other conversational products at scale. Inworld points to consumer apps on the platform that reached a million users in 19 days.
Best for Developers and product teams building realtime, consumer-facing voice and conversational AI..
Inworld fits teams whose roadmap turns on voice latency and cost per audio hour. If you are comfortable in code and want routing without markup, it is worth a close look. If you need a no-code tool, look elsewhere.
Pricing trap for Inworld AI: The headline TTS and STT rates apply on the Growth plan, so budget against your actual tier and expected usage.
Voice is where users decide if an agent feels real. Realtime TTS-2 starts speaking in under 100ms, and TTS-2 Flash in about 25ms, so replies land before a pause feels awkward. You can steer delivery with inline tags or free-form descriptions, clone or design voices, and carry one voice across 200+ languages.
Send speech in and get speech out over a single WebSocket or WebRTC connection. The assistant handles turn taking on its own, with adjustable eagerness, and you can register tools mid-session so it calls your functions without cutting the audio. Swap providers whenever latency, cost, or quality calls for it.
One API routes each request to the model that fits the moment, across OpenAI, Anthropic, Google, and 100+ others. You pay provider rates with no markup, and routing adds no latency. Failover, A/B testing, and model selection run on their own, so you can switch models without touching code.
Built-in analytics show whether your key metrics improve.
Realtime STT streams audio over a bidirectional WebSocket and returns a transcript as people talk. Each chunk carries voice profiling signals: emotion, age, accent, pitch, and style, so your app can shape its reply. Voice activity detection spots when speech starts and stops.
Word-level timestamps, diarization, and custom vocabulary support subtitles, search, and domain terms.
Inworld builds on a zero-trust framework with continuous monitoring, so safeguards run through the stack. It is SOC 2 Type II certified, HIPAA compliant, and GDPR compliant. That combination matters when procurement and legal review your stack before launch, and it lets you ship in regulated markets without adding extra vendors.
Build companions that feel present. OtherHalf uses Inworld to power voice-first companions at scale. Realtime TTS-2 answers in under 100ms, so replies arrive before users notice a pause.
Expected outcome: Replies that land fast in live conversation.
Route each request to the model that fits your latency, cost, or quality bar. Realtime STT streams live audio and reads emotion, age, accent, pitch, and style per chunk. Your agent adapts mid-call.
Expected outcome: Fewer drop-offs on live conversations.
Growth plan rates run $12.50 per million TTS characters and $0.10 per STT hour. LLM routing adds no markup, so you pay provider rates across 100+ models. > Tip: Track characters and audio hours as usage grows.
One WebSocket or WebRTC connection carries full-duplex audio. Register tools mid-session, and the assistant calls your functions without breaking the stream. Swap models anytime with no code changes.
Expected outcome: A working prototype with tidier wiring.
| Plan | Billed monthly | Billed annually | Includes |
|---|---|---|---|
Growth plan | See vendor page | See vendor page | Usage-metered: you pay per character synthesized and per hour of audio processed
|
Dedicated GPUs | See vendor page | See vendor page |
|
Tracking prices since Oct 9, 2026. No price change detected yet.
Pricing range: See vendor pricing
Pricing types: Monthly subscription, Usage-based pricing
Inworld AI prices its realtime voice stack by usage. The Growth plan bills realtime text-to-speech at $12.50 per 1M characters and realtime speech-to-text at $0.10 per hour, with zero markup on LLM routing and inference at 50% of the public rate.
Inworld AI prices its realtime voice platform on usage, with a Growth plan at $12.50 per 1M characters of realtime text-to-speech and $0.10 per hour of realtime speech-to-text.
Here's how those rates and the wider cost model break down.
Price: $12.50 per 1M characters (realtime TTS) and $0.10 per hour (realtime STT), billed by usage Best For: consumer-facing apps that want natural voice at scale without paying the list rates of other providers Other Features:
The Growth plan fits teams running live voice agents and companions, where a small per-unit saving adds up fast across thousands of sessions. You get the low latency and the routing flexibility without committing to a big contract. If you're still testing a single agent with a handful of users, the free-style tinkering stage matters more than volume rates, so start small and watch your character and audio-hour counts.
Price: from $5 per GPU-hour Best For: teams that need dedicated compute instead of shared infrastructure Other Features:
This tier is for workloads that have outgrown on-demand inference or need consistent performance under heavy concurrency. Smaller teams rarely need it, so hold off until your traffic justifies reserving hardware.
Worth knowing: the Growth rates above sit well under the public rates of comparable providers. Realtime TTS is quoted against $100 per 1M characters, and speech-to-text against $0.46 per hour, so the savings come from Inworld's standard list pricing rather than any promotion. Rates fall further as you move to enterprise scale, and LLM routing stays at zero markup no matter which provider model you choose.
Inworld doesn't publish a free tier or a money-back window alongside these rates. If you need to model spend before you commit, run a small pilot and measure characters and audio hours against your expected traffic — that gives you a realistic monthly figure far faster than guessing. Confirm license limits before you scale a commercial launch.
Email me when Inworld AI changes price
One email when we detect a price change on the vendor's pricing page. No newsletter, no deals. Unsubscribe in one click.
Voice features live or die on latency and cost. Inworld goes after both at once.
Other tools in the AI voice generator category
Vapi is an API-native platform built specifically for developers who need powerful voice AI capabilities. It’s the essential tool for creating sophisticated voice products and managing large-scale phone operations. The platform offers thousands of configurations and integrations, making the process simple and accessible. 💡 Teams from trail-blazing startups to Fortune 500 companies deploy their voice agents on Vapi.
Murf AI is an advanced platform focused entirely on streamlined AI voice technology. It provides a comprehensive AI Voice Studio for professionals who need to generate realistic voiceovers directly from text. You can choose from over 200 realistic voices and 10 different speaking styles. 💡 For developers, Murf also offers a suite of State-of-the-Art (SOTA) voice APIs. These APIs enable powerful capabilities like text-to-speech, voice cloning, and translation for integration into new products. Whether you are generating voices for corporate videos or designing conversational agents, Murf gives you fine-tuned control over output details like pitch, speed, tone, and specific pronunciations.
Insighto is an AI-powered Phone Agent Platform that automates customer conversations through AI-powered phone calls and chat, reducing wait times and improving business communications 🤖. It’s for businesses that want seamless, around-the-clock support across calls, SMS, email, WhatsApp, and live chat. It also offers fully white-labeled branding and easy CRM integrations with HubSpot, Zoho, Salesforce.
By usage. You pay per character synthesized and per hour of audio processed. There's no flat monthly price or per-seat fee, so budget against your real conversation volume and audio hours.
Inworld publishes no free tier and no money-back window. Confirm trial and refund terms with the team before you commit to spending, and pilot a small workload first.
You'll need API keys, a WebSocket or WebRTC connection, and your own orchestration. There's no drag-and-drop builder, so plan for real developer time on your first prototype.
The platform runs on a zero-trust framework and carries SOC 2 Type II certification, HIPAA compliance, and GDPR compliance, which helps clear procurement checks before you launch.
Are you Inworld AI? Claim this listing · Get the badge