Out of Set (@outofset) profile picture

We're building the most efficient voice AI.

Seoul
Joined September 2026
Photo by Out of Set (@outofset)Photo by Out of Set (@outofset)Photo by Out of Set (@outofset)
Out of Set retweeted
We re-benchmarked Airy TTS by @outofset after its text-processing update: it's now 2nd of 36 at reading numbers, dates and money in English (0.95, up from 0.50), 1st in Arabic, and the cheapest model on our board at $3.67 per 1M characters speko.com/benchmarks/tts
5
10
354
Out of Set retweeted
Airy Studioに日本語が追加されました🎉 > 無料でご利用いただけます > ボイスは300種類以上 👉 airy.so
1
1
38
Airy TTS is now on the @speko_ai benchmarks! It ranks #1 on cost and #8 on robustness among 31 models. The cost-robustness frontier now runs: Airy TTS → Gemini 3.8 Flash Lite → ElevenLabs v4 Turbo
3
54
Airy TTS is now the lowest-priced model on Coval! It sets a new Pareto frontier for cost vs. accuracy. @covaldev
1
1
5,067
Here’s a quick look at the last 7 days of results on Coval: 𝗖𝗼𝘀𝘁 > $3.67 per 1M characters > 59% cheaper than the next-cheapest model ($9) 𝗔𝗰𝗰𝘂𝗿𝗮𝗰𝘆 > 1.84% WER > Just 0.08pp behind the most accurate model (1.76%)
1
1
36
Airy TTS is here. > $2 per 1M characters, up to 96% cheaper > 0.2s end-to-end latency, up to 20x faster Most TTS APIs focus on time-to-first-audio. They're already fast there. But end-to-end latency is a different story. You're still waiting for the full response. Airy TTS solves that. You get a complete 27s audio clip back in just 0.2s. Up to 20x faster than other APIs. Try it now: airy.so
2
3
15,454