Emotion control
When you ask for a delivery, does the voice measurably change?
Ranking
Score out of 100, higher is better. 40 items per model; the thin line on each bar is the 95% range.
When two models' ranges overlap, this test cannot tell them apart.
Clip by clip
One square per clip: 10 sentences for each of four requested deliveries. Green means the voice measurably changed the right way against the model's own neutral reading.
Check on the method: models that offer no emotion control were measured the same way and passed 14 of 280 clips (5%). That is how often a green square appears by chance, because two takes of the same sentence never match exactly; scores of 10 or below mean "no control". This shows the delivery moved in the right direction; it does not prove a listener would name that exact emotion.
What we test and how it is scored
Test. 10 neutral sentences, each requested angry, excited, scared and whispered, through the model's own documented control (instruction field, inline tag or emotion setting). Models with no such control are rendered plain and scored the same way.
Scoring. Measured against the same model's neutral reading of the same sentence. Angry, excited and scared pass when median pitch rises by at least 2 semitones; a whisper passes when voiced sound drops by at least half. This checks that the delivery changed in the right direction. It does not prove a listener would name that exact emotion; blind listener voting is planned for that.
Full item list and rules: methodology.
Listen for yourself
Unedited clips from the test, in ranking order, including each model's worst miss where it had one.
Think a score is wrong? Check it.
Every test item, scoring rule and known limit is published, and the audio behind each number is on the page.