What are you trying to do?
I am trying to understand why audio/music generation using the YuE2 model on Maestro is taking significantly longer compared to running models on Wan2GP. For a standard song, Maestro took ~20 minutes to render, whereas Wan2GP finished a similar task in about 2 minutes. I’m looking to optimize Maestro’s render time or figure out if there is a VRAM bottleneck or offloading misconfiguration in my workflow.
What have you tried?
Switching execution back and forth between Maestro and Wan2GP to compare speeds, but the massive render time difference persists on Maestro.
Your system
OS: Windows 11
CPU: Intel Core i7-14700F
RAM: 32 GB DDR5
GPU: NVIDIA GeForce RTX 4060 (8 GB VRAM)
What are you trying to do?
I am trying to understand why audio/music generation using the YuE2 model on Maestro is taking significantly longer compared to running models on Wan2GP. For a standard song, Maestro took ~20 minutes to render, whereas Wan2GP finished a similar task in about 2 minutes. I’m looking to optimize Maestro’s render time or figure out if there is a VRAM bottleneck or offloading misconfiguration in my workflow.
What have you tried?
Switching execution back and forth between Maestro and Wan2GP to compare speeds, but the massive render time difference persists on Maestro.
Your system
OS: Windows 11
CPU: Intel Core i7-14700F
RAM: 32 GB DDR5
GPU: NVIDIA GeForce RTX 4060 (8 GB VRAM)