Left is faster. Higher reads more hard parts right.
Qwen3-TTS is better on naturalness and price. MAI-Voice-2.1 is better on non-streaming latency and accuracy.
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. Qwen3-TTS: 1.65 s. MAI-Voice-2.1: 861 ms.
MAI-Voice-2.1. Qwen3-TTS: 1.65 s. MAI-Voice-2.1: 861 ms.
Neither has a figure. Qwen3-TTS: Can't stream. MAI-Voice-2.1: Can't stream.
Neither has a figure. Qwen3-TTS: No stop message. MAI-Voice-2.1: No stop message.
MAI-Voice-2.1. Qwen3-TTS: 71.0%. MAI-Voice-2.1: 82.2%.
Qwen3-TTS. Qwen3-TTS: $20.00. MAI-Voice-2.1: $22.00.