The best of 18 voices at starting fast on streamed text
Streaming latency (median): −64 ms
Gradium TTS: the best of 18 voices at starting fast on streamed text. Its weakest is read from context, 24th of 26.
Last updated Methodology by Dileep Chagam
Where it ranks among all 26 voices, in plain words.
The first read of each test sentence, unedited.
Pick any voice to set against it. Every pair has its own page.
Gradium TTS starts in 76 ms on a typical sentence (median) and 105 ms on the slowest 1 in 10 (p90), 2nd of 26 voices.
Gradium TTS starts in 76 ms on a typical sentence (median) and 105 ms on the slowest 1 in 10 (p90), 2nd of 26 voices.
Yes. With streamed text, Gradium TTS starts 64 ms before the last word arrives.
Gradium TTS has no stop message, so the agent has to close the connection to stop it.
Gradium TTS read 79.5% of the hard parts right, as speech-to-text heard it. Its weakest kind is read from context, at 66.7%. This is an estimate, not a listening test.
Gradium TTS lists at $36.00 per million characters.