TURBO for ANIMA (less steps) Reviews

Thanks a lot for the update!

Been testing for a while now and v3 at 1.0 weight follows the prompt better than v1.5.

Caveats:

- I only tested manual prompts for less than 100 images, then the rest (~1000 images) were prompts that heavily rely on wildcards. I only have a few pose and outfit wildcards atm, that's why the examples in my two newest posts are repetitive.

- The only thing constant in my wildcard-based prompt is dutch angle and in v3 I got a lot more results in that angle compared to v1.5.

- Euler produces a lot of noise on my end, which only occurred on certain artist tags prior to using v1.5. And v1.5 pretty much eliminated the noise problem. So, I tested several sampler and scheduler combos and ended up with SA Solver + Simple. The noisy or oversaturated images in my post are the ones that didn't use SA Solver + Simple.

- I got a lot of overcooked images in v3 compared to v1.5 in my early tests, so I lowered the strength of the usual LoRAs I'm stacking.

I'll test inpainting, editing, outpainting, and USDU later - hopefully I won't need to change my parameters since I really like v3 more than v1.5 and would like to keep using v3. Thanks again :)