Left is faster. Higher reads more hard parts right.
Smallest Lightning v3.1 is better on streaming latency, accuracy and naturalness. MAI-Voice-2.1-Flash is better on non-streaming latency and price.
Last updated Methodology by Dileep Chagam
Measure by measure, the same sentences, the same place.
An estimate: speech-to-text transcribers scored it, not people.
The same sentence, read by each voice.
MAI-Voice-2.1-Flash. Smallest Lightning v3.1: 357 ms. MAI-Voice-2.1-Flash: 289 ms.
MAI-Voice-2.1-Flash. Smallest Lightning v3.1: 357 ms. MAI-Voice-2.1-Flash: 289 ms.
Smallest Lightning v3.1. Smallest Lightning v3.1: 363 ms. MAI-Voice-2.1-Flash: Can't stream.
Neither has a figure. Smallest Lightning v3.1: No stop message. MAI-Voice-2.1-Flash: No stop message.
Smallest Lightning v3.1. Smallest Lightning v3.1: 81.3%. MAI-Voice-2.1-Flash: 78.3%.
MAI-Voice-2.1-Flash. Smallest Lightning v3.1: $19.50. MAI-Voice-2.1-Flash: $15.00.