Inworld TTS Pricing at a Glance (October 2026)
Inworld TTS runs $15-$25 per million characters on-demand. The current lineup is two models: Realtime TTS-2 at $25/1M and Realtime TTS-2 Flash at $15/1M. A subscription drops both rates — the Creator plan ($25/mo) bills TTS-2 at $20/1M and Flash at $10/1M, and the discount deepens with each tier down to as low as $5/1M (TTS-2) and under $5/1M (Flash) at enterprise scale. Inworld cut these rates sharply in October 2026; the old $35/1M TTS-2 on-demand rate is gone, and the pay-as-you-go plan now includes about 70 minutes of free TTS to test with.
| Model | On-Demand | Creator ($25/mo) | Growth ($1,500/mo) | Enterprise |
|---|---|---|---|---|
| Realtime TTS-2 | $25/1M | $20/1M | $12.50/1M | As low as $5/1M |
| Realtime TTS-2 Flash | $15/1M | $10/1M | $7/1M | Under $5/1M |
TTS-2 vs TTS-2 Flash: which rate applies
TTS-2 ($25/1M) is the full closed-loop model — it listens to the conversation and adapts its tone in real time, which no other TTS service does. TTS-2 Flash ($15/1M) is the lower-latency, lower-cost variant for high-volume agents where you want most of that quality at 40% less. Both replaced the older TTS 1.5 Max/Mini line; if you still see the old $35/1M or "Founder rate" figures quoted elsewhere, they're out of date as of Inworld's October 2026 repricing.
Subscription Plans and Volume Discounts
Inworld uses a credit-based subscription model where monthly payments convert to API credits at discounted rates. Higher tiers unlock steeper per-character discounts:
| Plan | Monthly Cost | Monthly Credits | Effective TTS-2 Rate |
|---|---|---|---|
| On-Demand | $0 (pay-per-use) | ~70 min TTS free, then metered | $25/1M |
| Creator | $25/mo | $25 in credits | $20/1M |
| Builder | $100/mo | $100 in credits | Volume discount |
| Developer | $300/mo | $300 in credits | Volume discount |
| Growth | $1,500/mo | $1,500 in credits | $12.50/1M |
| Enterprise | Custom | Custom | As low as $5/1M |
The math vs ElevenLabs: Inworld TTS-2 is $25 per million characters on-demand, before any subscription discount. Compare that to ElevenLabs Scale — $299/mo for 1.8M credits on Multilingual v2 (1 credit ≈ 1 character), which works out to about $166 per million. That's roughly 6x more per character than Inworld's on-demand rate, and Inworld falls further on every tier down to the $5/1M enterprise floor.
What Makes TTS-2 Worth the Premium
TTS-2 launched May 5, 2026 as a "closed-loop voice model" — the first TTS that actually listens to the conversation before speaking. That sounds like marketing, but the architecture is genuinely different from every other TTS on the market:
- Conversational awareness. TTS-2 takes the actual audio of prior conversation turns as input, not just a transcript. It picks up the user's tone, pacing, and emotional state, then adjusts its delivery accordingly. The same line sounds different after a joke versus after bad news.
- Natural-language voice direction. Instead of preset emotion tags, you write prose descriptions like "tired but warm after a long day" and TTS-2 interprets them. Think of it as a system prompt for voice.
- Inline non-verbal cues. Five commands render as actual audio: [laugh], [breathe], [clear_throat], [sigh], [cough]. Drop them anywhere in text and they produce natural sounds — not the text-artifact versions most TTS services output.
- 100+ languages with voice identity preservation. Switch languages mid-sentence and the speaker still sounds like the same person. Cross-lingual voice cloning is built in.
- Three stability modes. Control the tradeoff between expressiveness and consistency depending on your use case — more predictable for IVR systems, more expressive for companions.
For the full technical analysis of how Inworld compares to competitors on quality, latency, and features, read our Inworld TTS review. It covers the prior TTS 1.5 Max generation in depth; TTS-2 builds on that foundation with the closed-loop architecture.
Voice Agent Economics: What TTS Actually Costs in Production
Inworld's sweet spot is voice agents — AI-powered phone bots, customer support, sales qualification. Here's what production deployments actually cost at Inworld's on-demand TTS-2 rate ($25/1M) versus ElevenLabs' Flash/Turbo API rate ($40/1M, i.e. $0.04/1K):
| Use Case | Daily Volume | Monthly Chars | Inworld TTS-2 | ElevenLabs Flash |
|---|---|---|---|---|
| Support bot (500 calls, 3 min avg) | 500 calls/day | ~11.3M | $283 | $452 |
| Sales qualifier (200 calls, 5 min avg) | 200 calls/day | ~7.5M | $188 | $300 |
| Virtual companion (8 hrs/day) | Continuous | ~18M | $450 | $720 |
| Contact center (10K calls, 4 min avg) | 10,000 calls/day | ~300M | ~$1,500 (Enterprise) | $12,000 |
The calculation: A typical 3-minute voice agent call generates about 750 characters of TTS output (the agent speaks roughly half the time, at ~125 words/min = ~625 chars + some padding). Monthly estimate: calls/day × 30 days × 750 chars.
At the enterprise contact center scale, Inworld's advantage is large: about $1,500/month at the ~$5/1M enterprise floor versus roughly $12,000 at ElevenLabs' $0.04/1K Flash API rate. That's close to an 87% cost reduction — and you get TTS-2's conversational awareness, which genuinely matters when your voice agent needs to respond to an upset customer differently than a casual inquiry. (Enterprise rates are negotiated; $5/1M is the published floor.)
Integration Costs: The Full Voice Agent Stack
TTS is just one piece of a voice agent. Inworld integrates with the major orchestration platforms, and each adds its own costs:
- Vapi — $0.05/min platform fee + telephony costs. The most popular voice agent platform. Inworld TTS is available as a provider option.
- LiveKit — Open-source with hosted option. No per-minute platform fee on self-hosted. Inworld TTS integrates via their Agents framework.
- Pipecat — Open-source Python framework by Daily.co. Free to use. Inworld is a supported TTS provider.
- NLX — Enterprise conversational AI. Custom pricing. Inworld is a partner for voice generation.
- Voximplant — Cloud communications platform. Per-minute telephony charges apply on top of TTS costs.
For the 500-call/day support bot scenario above, total monthly costs might look like: Inworld TTS ($339) + Vapi ($0.05/min × 500 calls × 6 min × 30 days = $4,500) + STT ($500) + LLM ($200) = ~$5,500. The TTS is typically 5-10% of total voice agent costs — the telephony and orchestration platform eat most of the budget. For STT pricing in voice agent stacks, see our Deepgram pricing guide.
Hidden Costs and Gotchas
- No ongoing free tier. Unlike ElevenLabs (10K credits/month forever) or Amazon Polly (5M chars/12 months), Inworld gives you about 70 minutes of free TTS to evaluate — then you pay. Good for testing, not for building a prototype.
- API-only — no consumer-facing product. There's no web editor, no drag-and-drop studio, no mobile app. You write code to use Inworld TTS. If you need a visual interface, look at Murf AI or ElevenLabs.
- Credits expire. Inworld's balance is dollar-denominated credits that roll over for up to three months, then lapse — you can't stockpile a year of unused credits. Match your plan to roughly what you'll actually spend each month.
- The Creator tier is a modest discount, not a big one. At $25/month, Creator bills TTS-2 at $20/1M versus $25/1M on-demand — a 20% cut. The real savings come at Growth ($1,500/mo, ~$12.50/1M) and Enterprise (as low as $5/1M). If your volume is small, on-demand is simpler.
- STT is separate. Inworld's speech-to-text (STT 1) is priced per hour of audio on top of TTS. For voice agents you need both — factor in both costs.
TTS-2 vs TTS-2 Flash: Which Model to Use
Inworld's October 2026 repricing collapsed the old TTS 1.5 Max/Mini line into two current models. The choice is now simpler than it used to be:
| Model | On-Demand Rate | Positioning |
|---|---|---|
| Realtime TTS-2 | $25/1M | Full closed-loop model — conversational awareness, natural-language voice direction, non-verbal cues ([laugh], [sigh]), 100+ languages. Use it when the agent responds to real users in real time. |
| Realtime TTS-2 Flash | $15/1M | Lower-latency, lower-cost variant of the same family — 40% cheaper. Use it for high-volume agents (IVR, notifications, basic phone bots) where you want most of the quality at a lower per-character price. |
My recommendation: Default to TTS-2 Flash for most agent workloads — at $15/1M it's already cheaper than ElevenLabs Flash ($40/1M) or Cartesia, and the quality gap to full TTS-2 is small for transactional conversations. Step up to full TTS-2 when the emotional read of the conversation is the product: healthcare, mental-health, retention, or companion apps where the model genuinely needs to respond differently to an upset user than a casual one. For exact per-model latency and language specs, check Inworld's docs, since those change faster than published pricing.
Inworld vs Competitors for Voice Agents
Voice agent builders care about three things: quality, latency, and cost. Here's how the options stack up:
| Service | Cost/1M Chars | Latency (P90) | Arena Rank | Emotion Awareness |
|---|---|---|---|---|
| Inworld TTS-2 | $15-$25 | <250ms | #9 (ELO ~1,266) | Yes (closed-loop) |
| Cartesia Sonic 3 | ~$37 | 40ms | #4 (ELO ~1,296) | No |
| ElevenLabs Flash | $40 (API) | ~300ms | #1 (v4, ELO ~1,338) | No |
| Fish Audio S2 Pro | $15 | ~150ms | Drama 3 #1 open-weight (~1,274) | 15,000+ emotion tags |
| OpenAI gpt-4o-mini-tts | ~$15 | ~300ms | Not ranked | Steerable via prompt |
| Gemini Flash TTS | ~$12 | ~250ms | #6 (ELO ~1,283) | 200+ audio tags |
| Grok TTS (beta) | $4.20 | ~400ms | Not ranked | No |
The tradeoff is clear: Cartesia wins on raw latency (40ms), but Inworld wins on quality (#1 ELO) and has the unique closed-loop awareness that no competitor matches. For voice agents where the AI needs to respond empathetically — think healthcare, mental health, customer retention calls — TTS-2's conversational awareness isn't a nice-to-have, it's the product differentiator.
For pure cost optimization where quality is secondary, Grok ($4.20/1M) or Gemini Flash (~$12/1M) are hard to beat. For a deeper speed-vs-quality analysis, see our Cartesia vs ElevenLabs comparison.
Who Should (and Shouldn't) Choose Inworld TTS
Pick Inworld if
- Building voice agents that need empathetic, adaptive responses
- Quality is the top priority (Arena #1)
- You need 100+ language support with voice identity
- You want non-verbal cues (laughter, sighs) in conversation
- Enterprise scale with negotiable rates
Skip Inworld if
- You need a web editor or studio UI (use ElevenLabs)
- Budget is the #1 factor (use Grok or Gemini)
- You need sub-50ms latency (use Cartesia)
- You want a free tier for ongoing use (use ElevenLabs or Chatterbox)
- You're creating audiobooks or pre-rendered content (the closed-loop advantage is wasted)
Related Pricing Guides
Use our TTS cost calculator to estimate Inworld costs at your specific volume. For the complete quality and feature analysis, read our full Inworld TTS review.
Comparing TTS options for a voice agent project? Our TTS API comparison guide ranks 12 APIs by latency, pricing, and developer experience. For a broader overview, see the best text-to-speech guide.
By TextToLab Research Team · Pricing verified October 16, 2026 against inworld.ai/pricing and Inworld TTS API docs. TTS-2 launched May 5, 2026. Arena rankings from Artificial Analysis.
Cost at your volume
Prices verified November 2026Enter how many characters of text you convert to speech each month and see what Inworld would cost versus the cheapest alternatives.
Inworld at 1,000,000 chars/month
Cheapest model per provider at 1,000,000 chars/month
| Provider | Model | $/1M chars | Your monthly cost |
|---|---|---|---|
| Amazon Pollycheapest | Standard | $4.00 | $4.00 |
| Google Cloud TTS | Standard | $4.00 | $4.00 |
| OpenAI | tts-1 | $15.00 | $15.00 |
| Azure AI Speech | Neural (S1) | $15.00 | $15.00 |
| Deepgram Aura | Aura-1 | $15.00 | $15.00 |
| Inworld | TTS-2 Flash | $15.00 | $15.00 |
| Fish Audio | s2.1-pro | $15.00 | $15.00 |
| ElevenLabs | v4 Turbo (API) | $40.00 | $40.00 |
| Cartesia | Sonic (Pro credits) | $50.00 | $50.00 |
Token-metered models (OpenAI gpt-4o-mini-tts, Gemini TTS) and subscription readers (Speechify) are billed differently and are not directly comparable per character — see each provider's page for details. Full cost calculator →
Sources — every price above was read from the provider's own pricing page
- Inworld: https://inworld.ai/pricing (verified 2026-11-05)
- OpenAI: https://developers.openai.com/api/docs/pricing (verified 2026-11-05)
- Amazon Polly: https://aws.amazon.com/polly/pricing/ (verified 2026-11-05)
- Azure AI Speech: https://prices.azure.com/api/retail/prices (verified 2026-11-05)
- Google Cloud TTS: https://cloud.google.com/text-to-speech/pricing (verified 2026-10-07)
- Gemini TTS: https://aistudio.google.com/docs/pricing (verified 2026-11-05)
- ElevenLabs: https://elevenlabs.io/pricing/api (verified 2026-11-05)
- Deepgram Aura: https://deepgram.com/pricing (verified 2026-11-05)
- Cartesia: https://cartesia.ai/pricing (verified 2026-11-05)
- Fish Audio: https://docs.fish.audio/developer-platform/models-pricing/pricing-and-rate-limits (verified 2026-11-05)
- Speechify: https://speechify.com/pricing/ (verified 2026-10-07)