SDXL Film Photography Style - CivArchive (CivitAI Archive)

SDXL Film Photography Style - beta v0.4

NSFW

READ EACH VERSION DESCRIPTION

Recommended settings:

1024 x 1024 resolution (most consistent)
Juggernaut XL (or similar photorealistic model)
- Use Natural Language prompts
0.8 Strength
No refiner
DPM++ 2M Karras
- >30 steps

Note: upscaling may degrade film grain

Trained on many high-resolution 35mm and medium format film images

The produced film grain may look bad if files are stored as .jpg, so .png is recommended

Most training images were shot with Portra, Ektar, or Superia X-TRA filmstock

Description

Better response to prompts, more coherent lines, and less jpeg artifacting than Alpha v0.2. Trained on a photorealistic model, not SDXL 1.0 base.

Use the following prompting template:

"film photography style" [your prompt], [modifiers], [shot type], [shot angle], [grain level]

[modifiers] - still a WIP, optional
- nothing, bokeh, bloom, god rays, moody, etc.
[shot type] - only use for portraits or pictures with people
- nothing, Extreme Close Up, Close Up, Medium Close Up, Medium Shot, Medium Full Shot, Full Shot
[shot angle] - describes the camera angle
- nothing, bokeh, bloom, god rays, moody, etc.
[grain level] - important to reduce Jpeg artifacts
- nothing, light grain, medium grain, heavy grain

** Note: nothing means "", so DO NOT type "nothing" **

FAQ

Comments (11)

lsed22998

Author

Oct 11, 2023· 1 reaction

CivitAI

I am going to train my second Beta model. More details to come

djcanOct 12, 2023· 1 reaction

CivitAI

Can I ask how many pictures are used for training? Thanks.

lsed22998

Author

Oct 13, 2023

Alpha 0.2 was ~100. Beta 0.4 was ~150

lsed22998

Author

Oct 13, 2023· 2 reactions

CivitAI

Beta v0.4 released! Let me know how it compares to Alpha v0.2

dioOct 29, 2023

CivitAI

The grain and grade are lovely but it seems to have a serious sameface+frowny face issue

StephanTualDec 29, 2023

CivitAI

Author: first this is brilliant I'm glad fooocus uses it as default on 'realistic' mode - but is it expected for it to 'rethink' the images based on the same seed as oppose to 'refining' them? I get major differences (just tries a batch of 100- all are consitently quite different from the original). Thank you!

antoinevdlb422Mar 7, 2024· 1 reaction

CivitAI

Anyone know why this isn't appearing in the LORA list after dropping it in the right file?

3955840May 4, 2024· 2 reactions

CivitAI

beta 4.0 makes faces lop-sided and skewed among other serious errors

lsed22998

Author

May 20, 2024· 1 reaction

I will be releasing the next version very soon, and I hope it will better resolve this issue!!

3955840May 20, 2024· 1 reaction

@lsed22998 sounds great because it has lots of other great qualities, im unsure why it would skew the faces like that though. its a very interesting curiosity haha

lsed22998

Author

May 20, 2024

@MikeLitorous Thank you! I just uploaded the first release, so please let me know how those faces are!

LORA

SDXL 1.0

by lsed22998

Download (Beta) View on CivitAI