Inference speed

Realtime speed is total generated audio duration / total request time; higher is better. 3.3x means 3.3 seconds of audio generated per second. RTF is request time / audio duration; lower is better. TTFA is measured only for audio events.

Run Engine Checkpoint Delivery Completed Realtime speed (×) Median RTF p95 RTF Median client TTFA ms TTS errors
20261025-050404-08222176 faster 0.6B-CustomVoice complete 100/100 3.30 0.303 0.319 n/a 0
20261025-050404-08222176 faster 0.6B-Base complete 30/30 3.29 0.305 0.316 n/a 0
20261025-053632-9d5b8428 faster 1.7B-CustomVoice complete 100/100 2.88 0.348 0.369 n/a 0
20261025-053632-9d5b8428 faster 1.7B-Base complete 30/30 2.94 0.339 0.357 n/a 0
20261025-053632-9d5b8428 faster 1.7B-VoiceDesign complete 100/100 2.86 0.348 0.376 n/a 0

See DETAILS.md for the latest run and methodology.