Model changes: Replaced Turbo LoRA v0.2 with the new turbo-v1.0 checkpoint . This has significantly improved textures but slightly weaker aesthetics. A small amount of the previous Turbo LoRA was merged back in to help stabilize the image. Refined my approach to block pruning for smoother mathematical curve. In theory, this should reduce distortions caused by sharp changes across block...
Model changes:
Replaced Turbo LoRA v0.2 with the new turbo-v1.0 checkpoint . This has significantly improved textures but slightly weaker aesthetics.
A small amount of the previous Turbo LoRA was merged back in to help stabilize the image.
Refined my approach to block pruning for smoother mathematical curve. In theory, this should reduce distortions caused by sharp changes across block strengths.
Added a splash of Blocks 11 and 12 from my custom training as they help tame colors without radically affecting composition.
Workflow changes:
Switched from qwen_image_vae to wan_2.1_vae for slightly better microdetail.
INT8 conversion path now uses the convrot outlier method for improved quality.
New custom sigma curve. It is just beta scheduler with the first value bumped slightly above 1.0. This is an easy way to add detail.
Slight adjustments to default fluff prompt.
Other notes:
Previous versions of Photanima converged at 6 steps, but v2.3 needs at least 8 steps.
Model changes: Replaced Turbo LoRA v0.2 with the new turbo-v1.0 checkpoint . This has significantly improved textures but slightly weaker aesthetics. A small amount of the previous Turbo LoRA was merged back in to help stabilize the image. Refined my…
Model changes:
Replaced Turbo LoRA v0.2 with the new turbo-v1.0 checkpoint . This has significantly improved textures but slightly weaker aesthetics.
A small amount of the previous Turbo LoRA was merged back in to help stabilize the image.
Refined my…
