whatmodel.sh / tts / Inworld Realtime TTS-2

Inworld Realtime TTS-2

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Inworld Realtime TTS-2 is a closed-weights text-to-speech model from Inworld, released August 2026. Licence: proprietary.

61.0±1.8#39 of 118 text-to-speech modelsfrom 7 benchmark results across 4 sources

Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.

$ how it compares

The text-to-speech models ranked either side of Inworld Realtime TTS-2, out of 148 with a score, then more from Inworld.

#modelscorewhereRAMreleased
46DMOSpeech 2 Columbia University / NewsBreak (Yinghao Aaron Li, Xilin Jiang et al.)62±4local~4.3GB2025-07
47Eleven Turbo v2 ElevenLabs61±3api $50.00/M chars2023-11
48Inworld Realtime TTS-2 Inworld61±2api $25.00/M chars2026-08
49CosyVoice2-0.5B Alibaba (FunAudioLLM)61±1local~4.1GB2024-12
50E2 TTS Base (F5-TTS reproduction) SWivid (SJTU X-LANCE)60±2local~1.3GB2024-10
more from Inworld
41Inworld TTS-1.5 Max Inworld62±3api $10.00/M chars2026-01

see all text-to-speech models · find the ones that fit your machine

$ what it's good at

other tests7 not in a task · 99% of its score
testresultvs the others in its rundifficultycountswhy
Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot1197.8 ✎▁···▁▃··█▄▁▇ beats 8/18avg 1078▇▇▇▇▇▇▇▇▇▇▇▇ 29%typical
Vapi Humanness · historical Elo, 2026-06-101189.4 ✎▃·▃▃▃·▃██▃·▃ beats 4/12avg 1080▇▇▇▇▇▇ 15%typical
ZTTS1-Eval in-the-wild · 10-language mean WER3.8% ✎▄·····█··▄██ beats 1/7avg 8.3%▇▇▇▇▇ 13%typical
ZTTS1-Eval Clean (FLEURS-R) · 9-language mean WER4.5% ✎▄·····▄█▄█·▄ beats 3/7avg 12%▇▇▇▇▇ 12%typical
ZTTS1-Eval in-the-wild · 10-language mean MSR-UTMOS2.81 / 5 ✎█··▄█▄····▄▄ beats 4/7hard · avg 2.00 / 5▇▇▇▇ 11%at its level
ZTTS1-Eval Clean (FLEURS-R) · 9-language mean MSR-UTMOS3.18 / 5 ✎▃▃█···▃···▃▃ beats 3/7medium · avg 2.52 / 5▇▇▇▇ 10%at its level
RW-Voice-EQ TTS · Acoustic Quality (human)3.33 / 5 ✎█·█········▄ beats 1/4medium · avg 2.70 / 5▇▇▇ 8.2%at its level

$ where to run it

Through an API

api · Replicate · $25.00/M chars · replicate.com ↗ · price sourceUSD per 1M input characters. Realtime TTS 2.0, standard hosted endpoint. Character billing, not a token conversion. Replicate output/character billing; hardware-time rates are not added to this tariff. Taxes and separate services excluded. Bertrande, https://bertrande.com/en/ai/replicate/inworld--realtime-tts-2; observed 2026-09-20T01:44:15Z. CC BY 4.0; units normalized and model identifier mapped.

$ how the score was worked out · ZONOS2 Technical Report (ZTTS… 46%, The Humanness Index by Vapi ·… 29%, The Humanness Index by Vapi ·… 15% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
ZONOS2 Technical Report (ZTTS1-Eval results)4▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 46%CC BY-SA 4.0
The Humanness Index by Vapi · September 2026 public standings1▇▇▇▇▇▇▇▇▇ 29%CC BY 4.0
The Humanness Index by Vapi · June 2026 production export1▇▇▇▇▇ 15%CC BY 4.0
RW-Voice-EQ Bench (Hume AI)1▇▇ 8%CC BY 4.0
prior · pull toward the middle▇ 1%
weighted average= 61.0

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json