Area 08 · 4 jobs

Best AI for audio and voice, job by job

We don't rank models in general. We test them one job at a time: the same inputs and settings for every model, graded blind, with the cost of each usable result. 1 of these jobs are being tested now.

Testing now

Voiceover and text-to-speech, multilingual

Testing now
Why it's hard

Robotic, wrong tones in Chinese, bad pauses

What we test

Same 3 scripts in EN, 中文, JA, KO

Coming in month 2

Voice cloning

Month 2
Why it's hard

Clone sounds off after a few sentences

What we test

Consented 30 s sample → 2 min read

Music and background tracks

Month 2
Why it's hard

Copyright worry; generic tracks

What we test

5 briefs, licence terms checked

Transcription and subtitles (中英混合)

Month 2
Why it's hard

Mixed Chinese/English mangled

What we test

3 recordings, error rate

How we choose and grade: our methodology.