Left is faster. Higher reads more hard parts right.
Inworld TTS-2 is better on non-streaming latency, streaming latency, accuracy and naturalness. AssemblyAI TTS 1 is better on stopping 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.
Inworld TTS-2. Inworld TTS-2: 126 ms. AssemblyAI TTS 1: 171 ms.
Inworld TTS-2. Inworld TTS-2: 126 ms. AssemblyAI TTS 1: 171 ms.
Inworld TTS-2. Inworld TTS-2: 127 ms. AssemblyAI TTS 1: 159 ms.
AssemblyAI TTS 1. Inworld TTS-2: No stop message. AssemblyAI TTS 1: 40 ms.
Inworld TTS-2. Inworld TTS-2: 93.4%. AssemblyAI TTS 1: 81.5%.
AssemblyAI TTS 1. Inworld TTS-2: $25.00. AssemblyAI TTS 1: $15.00.
Cekura runs these simulated calls on your agent and scores every one.