too slow..

[ssadae1029]

On an RTX 5090, it takes a full 13 minutes to generate a 25-second audio clip with Step-Audio-EditX TTS.
Is this normal? It feels insanely slow compared to what I expected from a 5090.
For reference: 25 seconds of audio → 13 minutes of generation time (RTF ≈ 31x).

[Sweet-john]

If it’s really like that, I won’t consider using it ...

[nvnhat95]

It's indeed slow but not that slow. RTF is around 4 on A5000.

[ssadae1029]

It's indeed slow but not that slow. RTF is around 4 on A5000.
I reconfigured everything and ran a bunch of tests.
The moment it goes over 3 sentences, accuracy drops hard and the pronunciation gets completely messed up.
Like you said, I'm getting about 3x the generation time... honestly way more disappointed than I expected...