Ready.
The first request can include ZeroGPU queue and model-transfer time. Use the audio player's menu to download the final WAV.
The same text, seed, speed, variation, and pitch are sent to both models.