I’ve seen a lot of people complaining about V7 being slow to generate images and throwing tons of errors, but honestly, a lot of that isn’t actual bugs. The whole architecture of V7 is different—queue, upscaling pipeline, parameter behavior—all changed from V6. From what I’ve tested, the most common issue is getting stuck at “waiting to start” in the queue.
Sixty to 120 seconds is totally normal, and it gets worse during peak hours from afternoon to evening. That’s not a freeze—it’s just that the diffusion model itself is heavier and processes more per image. If you have Fast credits, switching to Fast mode helps.
And a bunch of those “errors” are just people still using old V6 prompt habits that don’t translate. Trying to force those will naturally mess up. Don’t just jump in and blame it—first figure out if it’s a real failure or you’re just using it like the old version.