5 ms·

It's impressive that it spell words correctly and lay them out but the issue I have is the text always has this distinctively overly fried look to it. The color

It's impressive that it spell words correctly and lay them out but the issue I have is the text always has this distinctively overly fried look to it. The color of the text is always ramped up to a single value which when placed into a high fidelity image gives the impression of just slapping some text on top with photoshop afterwards in quite an amateurish fashion rather than text properly integrated into an image.

imiric 3y ago
But the sample images they show here showcase a good job at blending text with the rest of the image, using the correct art style, composition, shading and perspective. It seems like an improvement, no?
- blehn 3y ago
  The blending looks better, but the LED sign on the bus for example looks almost like handwritten lettering... the letters are all different heights and widths. Not even close to realistic. There's a lot of nuance that goes into getting these things right. It seems like it'll be stuck in an uncanny valley for a long time.
viraptor 3y ago
It's just the presented examples. See the first preview for examples of properly integrated text https://stability.ai/news/stable-diffusion-3 https://stability.ai/news/stable-diffusion-3 Especially the side of the bus.
- whywhywhywhy 3y ago
  "Go" "Stable Diffusion" are not they look slapped on and would be sent back in most studios. Dream on is a bit better but looks very messy, badly spaced, inconsistent letter sizes.
- MyFirstSass 3y ago
  Side of the bus still looks weird though. Like some lower resolution layer was transformed to the side, also still to bright. Still impressive we've come to this but we're now in the uncanny valley which is a badge of honour tbh.
bsenftner 3y ago
I'm expecting at some point the stable diffusion community of developers to recognize the value of Layered Diffusion, the method of generating elements with transparent backgrounds, and transitioning to outputs that are layered images one may access and tweak independently. The addition of that would make the hands-on media producers of the world say "okay, now we're talking, finally directly indigestible into our existing production pipelines."
- cthalupa 3y ago
  There's already ComfyUI nodes for Layered Diffusion. https://github.com/huchenlei/ComfyUI-layerdiffuse https://github.com/huchenlei/ComfyUI-layerdiffuse Of the people I know in the CG industries using SD in any sort of pipeline, they're all using Comfy because a node based workflow is what they're used to from things like Substance Designer, Houdini, Nuke, Blender, etc.
  - bsenftner 3y ago
    That's my impression as well. I used to work in VFX, my entire career is pretty much 3D something or other, over and over.
GaggiX 3y ago
It's very likely an artifact of CFG (classifier-free guidance), hopefully some days will be able to ditch this kinda dubious trick. This is also the reason why the generated images have this characteristic high contrast and saturation. Better models usually need to rely less on CFG to generate coherent images because they fit the training distribution better.

imiric 3y ago

But the sample images they show here showcase a good job at blending text with the rest of the image, using the correct art style, composition, shading and perspective. It seems like an improvement, no?

blehn 3y ago
The blending looks better, but the LED sign on the bus for example looks almost like handwritten lettering... the letters are all different heights and widths. Not even close to realistic. There's a lot of nuance that goes into getting these things right. It seems like it'll be stuck in an uncanny valley for a long time.

viraptor 3y ago

It's just the presented examples. See the first preview for examples of properly integrated text https://stability.ai/news/stable-diffusion-3 https://stability.ai/news/stable-diffusion-3 Especially the side of the bus.

whywhywhywhy 3y ago
"Go" "Stable Diffusion" are not they look slapped on and would be sent back in most studios. Dream on is a bit better but looks very messy, badly spaced, inconsistent letter sizes.
MyFirstSass 3y ago
Side of the bus still looks weird though. Like some lower resolution layer was transformed to the side, also still to bright. Still impressive we've come to this but we're now in the uncanny valley which is a badge of honour tbh.

bsenftner 3y ago

I'm expecting at some point the stable diffusion community of developers to recognize the value of Layered Diffusion, the method of generating elements with transparent backgrounds, and transitioning to outputs that are layered images one may access and tweak independently. The addition of that would make the hands-on media producers of the world say "okay, now we're talking, finally directly indigestible into our existing production pipelines."

cthalupa 3y ago
There's already ComfyUI nodes for Layered Diffusion. https://github.com/huchenlei/ComfyUI-layerdiffuse https://github.com/huchenlei/ComfyUI-layerdiffuse Of the people I know in the CG industries using SD in any sort of pipeline, they're all using Comfy because a node based workflow is what they're used to from things like Substance Designer, Houdini, Nuke, Blender, etc.
- bsenftner 3y ago
  That's my impression as well. I used to work in VFX, my entire career is pretty much 3D something or other, over and over.

GaggiX 3y ago

It's very likely an artifact of CFG (classifier-free guidance), hopefully some days will be able to ditch this kinda dubious trick. This is also the reason why the generated images have this characteristic high contrast and saturation. Better models usually need to rely less on CFG to generate coherent images because they fit the training distribution better.