Bracketed cue
Why we chose this prompt: Bracketed stage directions like [whispering] and [excited] are a common way to direct delivery. The model has to act on them and not read them aloud, and the two cues in one prompt test whether it can switch.
[whispering] Route 6 Espresso is open till nine, and the last train leaves at eleven fifty-six. [excited] The last train leaves at eleven fifty-six!
provider default
Cost--Generation time27.1sCorrectness100%george
Cost--Generation time2.7sCorrectness100%george
Cost--Generation time2.6sCorrectness100%george
Cost--Generation time2.8sCorrectness100%provider default
Cost--Generation time3.3sCorrectness100%provider default
Cost$0.002Generation time1.5sCorrectness100%
Zephyr
Cost--Generation time10.0sCorrectness100%Zephyr
Cost--Generation time5.6sCorrectness100%
en-US-Harper:MAI-Voice-2
Cost$0.003Generation time2.3sCorrectness100%
loongjohn
Cost$0.002Generation time3.1sCorrectness100%longanlingxin
Cost$0.003Generation time2.9sCorrectness100%nadine
Cost--Generation time3.5sCorrectness96%nadine
Cost--Generation time13.6sCorrectness96%george
Cost--Generation time4.4sCorrectness93%provider default
Cost$0.002Generation time2.3sCorrectness93%Zephyr
Cost--Generation time5.2sCorrectness93%eve
Cost$0.002Generation time1.6sCorrectness93%en-US-Harper:MAI-Voice-2
Cost$0.002Generation time1.9sCorrectness90%
en_paul_sad
Cost$0.000Generation time2.1sCorrectness90%tara
Cost$0.001Generation time10.5sCorrectness86%george
Cost--Generation time0.7sCorrectness86%provider default
Cost--Generation time1.3sCorrectness86%provider default
Cost--Generation time2.6sCorrectness83%
aura-2-thalia-en
Cost$0.004Generation time4.1sCorrectness83%george
Cost--Generation time1.1sCorrectness83%george
Cost--Generation time2.4sCorrectness83%george
Cost--Generation time1.9sCorrectness83%george
Cost--Generation time0.7sCorrectness83%English_expressive_narrator
Cost$0.015Generation time2.3sCorrectness83%English_expressive_narrator
Cost$0.009Generation time1.4sCorrectness83%
af_alloy
Cost$0.000092Generation time3.6sCorrectness77%flux-alexis-en
Cost--Generation time9.0sCorrectness74%conversational_a
Cost$0.001Generation time9.9sCorrectness44%voice_3150
Cost--Generation time9.1sCorrectness4%geffen_32
Cost--Generation time1.7sCorrectness--