Inworld · read on October 27, 2026
Inworld Realtime TTS-2
Inworld Realtime TTS-2 is a TTS (text-to-speech) model from Inworld. Coval measures its P95 time to first audio at 194.7 ms, which puts it in the Fast enough speed tier. It ranks 8 of 86 in the full ranking.
- Speed tier
- Fast enough
- P95 / median
- 194.7 ms / 149.1 ms
- Listener Elo
- 1,267
- Price per 1M chars
- $15 to $25
Where it ranks
Full ranking: 8 of 86. Use case ranks, by speed:
- Customer service: 5 of 70
- Voice agents: 8 of 13
- Assistants: 5 of 69
- Entertainment: 5 of 69
- Knowledge sharing: 5 of 73
The tier and the ranks are computed by this site from the benchmarks' figures, not declared by Inworld.
Measurements and sources
Coval
| Median time to first audio (P50) | 149.1 ms |
| Time to first audio, P95 | 194.7 ms |
| Time to first audio, mean | 165.8 ms |
| Time to first audio, P90 | 181.9 ms |
| Timed requests | 2,612 |
Speko
| Naturalness score | 1,561 |
| Latency, board chart | 84 ms |
| Latency, model page sweep (median) | 116 ms |
| Numbers and dates pronunciation | 0.61 |
Artificial Analysis
| Listener Elo | 1,267 |
| Listener Elo, 95% interval | 14 |
| Listener Elo, customer service | 1,311 |
| Hard passages pronounced correctly | 80.7% |
Cekura
| Time to first audio, streamed text | 127 ms |
| Time to first audio, text sent whole | 126 ms |
| Reading accuracy | 93.4% |
| Estimated naturalness (1 to 5) | 4.52 |
Details
| Provider | Inworld |
| Access | Proprietary |
| Streaming | A benchmark measures its time to first audio |
| Features listed by Coval | emotion-control, multilingual, voice-cloning |
| Languages tested | Spanish, German, French, Portuguese, Japanese, Chinese, Arabic, Hindi, Vietnamese, Korean, Filipino, Thai |
| Measured by | Artificial Analysis, Cekura, Coval, Speko |
| Official site | inworld.ai |
Sources
- https://artificialanalysis.ai/text-to-speech (read on October 27, 2026)
- https://artificialanalysis.ai/text-to-speech/leaderboard/provider-voice (read on October 27, 2026)
- https://benchmarks.cekura.ai/text-to-speech (read on October 27, 2026)
- https://benchmarks-api-6wxgp27p2a-ue.a.run.app/v1/results/aggregates?benchmark=TTS&window=30d&include_series=false (read on October 27, 2026)
- https://benchmarks.speko.ai/tts/ (read on October 27, 2026)
Values are copied from these pages on the date shown. The match between this model and its benchmark entries is inferred, because a source writes the name differently: Cekura: “Inworld TTS-2”; Coval: “Inworld AI TTS 2”; Speko: “Inworld tts 2”. See the methodology.
Compare it with the reference model: Gradium TTS vs Inworld Realtime TTS-2.