Civit's generator not letting you use a negative value? Use v1.0-negated, which makes it so positive values zoom out!
Zooms in and out on character portraits. I've tried to balance it so that 1.0 is zoomed in close to the character's face, and -1.0 is a full body shot, however, your results may vary depending on your subject and what checkpoint you're using. I've had generally good luck up to +/-1.25 and even more in some cases.
This works with both anime and photorealistic generations. It was trained with humans in mind, and has not been tested with animals or scenery, so YMMV.
Description
FAQ
Comments (56)
This doesn't show up in my LORA list after uploading it into the folder for some reason
+1, we need some help.
For the record, it doesn't show up in mine either. I always just type the names of LoRAs, so I didn't even notice.
I'd report this as a bug to automatic1111.
For whatever reason, it shows up for me in Auto111 WebUI. I did make sure to have an SDXL checkpoint loaded first.
@_Envy_ It does show up for me, buuuut the civitai extension didn't calculate a hash for it so it wasn't cross-posted like it should have. I remember that being the case for lycos, but this isn't one, is it?
Works great but doesnt show in my Lora list.
Can invoke using <lora:EnvyZoomSliderXL01:1.25>
This doesn't show up in LoRA in A111.
In the settings, check the "Always show all networks on the Lora page" in the Extra networks section.
@TonyZed68 Thanks, that fixed it for me.
@TonyZed68 Thanks ! This worked for me :)
IT'LL SHOW UP in search when you're USING >>> COPAX TIMELESSXL - SD ...
or NewRealityXL (for example) ^^
p.s.
i'm so happy - thank you Envy !!
For anyone having trouble making it appear in the LoRA section, use a model that isn't sdxl, go to LoRA section, click on the hammer and wrench icon in the top right corner (refresh if need be), and under "stable diffusion version" just choose SDXL
Thanks for doing this!
Can you do me a favor and tell me if this lora is doing the same thing?
@_Envy_ It appears like it is, the same fix does solve the issue though. The default seems to be an older version like 1.5 or in this case 1.0.
Thank you :)
Thanks for this. I thought I'd done something daft again!
Any chance you'll retrain this to work with Pony? I haven't managed to make the slider training work myself.
I've found that this works fairly well with Pony realism models.
@Suppressor sadly not for anime ones, like autismmix, which I'm sure would make it work for most other pony.
@lost_moon I'm not sure if it's even possible. Since I made this post I have managed to get slider training to work, but the zoom slider does nothing after I've trained it.
@FatsIncognito Didn't work for me either.
If you want to train a slider, you need a checkpoint that has at least some inkling of the concept you want to train in, and Pony's comprehension of anything other than booru tags has been beaten out of it. I suspect that its complete lack of understanding about zooming in and out is why this lora doesn't work on it.
@_Envy_ That would explain why the realism models with SDXL infusions can vaguely comprehend it
I think this LoRa is extremely good. Good work!!!!
Anything for SD 1.5?
https://civitai.com/models/111292?modelVersionId=120059
This does almost the same for 1.5 ;)
thanks for sharing. I dont know if is Forge the problem but doesnt seems to make any effect when i apply the lora.. will try on auto
Amazing LoRA! Thanks!
This lora configuration fails in both 1.5 and xl models, as does adjusting weights, changes, does not work on the screen, and switching large models does not work
Well, it's an XL model so it's not going to work in 1.5. What software are you using? Can you post an attempt?
@_Envy_ Hi there! I'm having this issue too on an XL-based model. I posted an attempt (showing prompts and everything) to my profile, its the blonde maid girl.
This model doesn't work with Pony Diffusion. The model you use has to have some concept of what zooming means, and pretty much everything that's not a danbooru tag was beaten out of Pony during training (which unfortunately makes it impossible for me to use the same method to make a Pony version as well).
@_Envy_ Ohhh I see. Thank you!
this is really a nice hack!
you don't how many times I get annoyed by sd not setting up the camera distance.
what annoys me the most is it keep changing for every change in prompt/seed. But still it is really a good idea to have more control!
I hear that. ED is like trying to get a belligerent child to listen to you when they're having a temper tantrum. Like, zoom out, show room, view outside room, etc. ect. ect., and it still does a closeup shot of one character when there's supposed to be three in the room but there's only one and a half cuz one is stuck in a counter..
Now we really need this for SD3!
I have a problem where my gens ignore shot composition and like close ups, I'm starting to think it's the models that I use. See here for an example, there's essentially no difference https://civitai.com/posts/3553771
I can get this lora to work -> https://civitai.com/posts/3668833
Edit: We need the same for flux: YOU CAN DO IT !
Ahhhh i miss him so much on flux!
Old message: Lora very helpful (in SDXL). I always have it ready in my lora stack if needed ;-)
Any chance you could make a Zoom Slider for Flux? It is badly needed. I can donate Buzz if that helps?
Don't need a donation. I'll give it a shot. :)
@_Envy_ sounds great! it's badly needed - good luck:-)
I second this idea. I've gotten more use out of the SDXL zoom slider than any other lora. I produce one-off artwork for guitar effects pedals I build, and this really helps sort out placement. I'm just experimenting with Flux and I was thinking: I really need to adjust the zoom.
@_Envy_ i hope you succeed, i always prefer Flux over XL now
Absolutely love this LoRA! Very well done my friend! Tipping 1000 buzz to show appreciation! Please keep up the great work!
Thanks!! Glad it's working for you!
@_Envy_ something changed with the Civitai generation, no longer can do negative on the slider... ?
@Papa_bradley I uploaded a negated version of the LoRA where higher numbers zoom out:
https://civitai.com/models/213307?modelVersionId=949771
Otherwise, it's exactly the same LoRA.
@_Envy_ Great! Thanks!!
Could you tweak this to allow a wider range of variables on Civitai? Right now it's limited to -1, but being able to zoom out further would be helpful.
Please ping me if you're able to do that, I got some buzz to send your way if so.
I changed the minimum and maximum strength, but I have a feeling this is limitation is a civitai issue.
I've uploaded a negated version of this LoRA here:
https://civitai.com/models/213307?modelVersionId=949771
It's exactly the same as this one except positive numbers zoom out and negative numbers zoom in.
@_Envy_ sweet, look forward to experimenting with it
Details
Files
EnvyZoomSliderXL01.safetensors
Mirrors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01_EnvyZoomSliderX.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
EnvyZoomSliderXL01.safetensors
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.