First of all: thank you so much @thibaudart for sharing such a wonderful workflow with the community <3
The first run did a pretty amazing job for me – very happy:
Since my Mac took quite a long time for one 4K upscale (23 minutes), I wanted to get some insights before the next run. In a few places I could spot some digital artifacts / patterns as you can see here:
Is there a setting / setup, which can prevent such artifacts or is it just "hit and miss" :)? My settings were …
model = turbovisionxlSuperFastXLBasedOnNew_tvxlV431Bakedvae.safetensors
steps = 10
cfg = 5
sampler = euler
scheduler = karras
Input resolution = 896 x 1152 px
scale_by = 0.5
Upscale = 2.00
Hallucination = 0.2
Tile = 1.0
Bonus question: in what way does the "Tile Size"-value influence the generation?
First of all: thank you so much @thibaudart for sharing such a wonderful workflow with the community <3
The first run did a pretty amazing job for me – very happy:
![2](https://github.com/fictions-ai/sharing-is-caring/assets/136800027/601f637a-f753-42fb-9481-36c6bd54dce4)
Since my Mac took quite a long time for one 4K upscale (23 minutes), I wanted to get some insights before the next run. In a few places I could spot some digital artifacts / patterns as you can see here:![3](https://github.com/fictions-ai/sharing-is-caring/assets/136800027/e349e15d-9505-490a-87d4-3283246bf32c)
Is there a setting / setup, which can prevent such artifacts or is it just "hit and miss" :)? My settings were …
Bonus question: in what way does the "Tile Size"-value influence the generation?
Cheers