https://civitai.red/models/155444/yafm-yet-another-f-merge
Pro
18
YAFM: Yet Another F Merge
159
1.6k
49
Updated: Jun 11, 2025
base model
photorealistic
lora-friendly
realism
Remix
10657-338187537-Heun-7-40-cybermix last.fp16.jpg
Remix
10645-978827126-Heun-7-40-cybermix last.fp16.jpg
Images hidden due to mature content settings
Download
1 variant available
fp16 SafeTensor
Half precision, best balance • 4.87 GB
Verified: 3 years ago
Details
Type
Checkpoint Merge
Stats
1,343
0
1
2
3
4
5
6
7
8
9
,
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
Reviews
Very Positive
(150)
Published
Oct 2, 2023
Base Model
SD 1.5
Hash
AIR
civitai:
155444
@
174359
default creator card background decoration
Followers - 12
12
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
Likes - 159
159
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
charless's Avatar
charless
Joined May 3, 2023
License:
CreativeML Open RAIL-M
Addendum
Hide
EDIT For V2:
Continued testing/merging/going forward then backward then forward, tweaking etc.
Felt like sharing a new version after all this time.
I like it better than the old YAFM, that's the reason I'm sharing it.
For v2 illustration:
I basicaly just used my wildcards to generate a group of pics. Grabbed everything (except 2 of them that where really awful) and posted all of them here. Made this to avoid cherrypicking.
So if the pics aren't great, that's normal.
V1:
I'll be quick, I had only 2 goals merging this thing:
it will be somewhat realistic
it will use my lora/lyco and not render awful (or not if you like them) Mars Mutants Imported from Verhoeven's Mind.
(I'm using it everyday so I guess it didn't fall short of my expectations)
It's been posted on civitai because someone asked for it. I'm not providing any update or anything. Just the merge as it is now.
So basically, it can render your favorite girl/celebs lora/lyco. I trained mine using the standard sd-v1-5-pruned-noema-fp16.safetensors and it's doing well. I don't know for others.
Now, all the Cheers and Kudos and Love and Kisses go to the original models creator of checkpoints listed here:
cyberrealistic_v31
https://civitai.com/models/15003?modelVersionId=89680
fantasticmix_v40
https://civitai.com/models/22402?modelVersionId=59385
almosthumanMix_v20
https://civitai.com/models/107132/almosthuman-mix
epicdream_lullaby
https://civitai.com/models/125189?modelVersionId=136739
photon_v1
https://civitai.com/models/84728?modelVersionId=90072
Concerning the illustrative pics:
It was quick gens, between 1-3 tries for each one. I didn't want to spend hours generating them. I don't see any point cherry picking images or posting inpainted one. It's all '1 click on generate and I wait' pics.
Concerning Prompts:
Positive:
Keep them simple, cf what I posted here. You can see some more examples of what i call simple in photon_v1 creator's gens.
(photo portrait:1.0), a woman wearing tennis top and shorts, smiling, brown hair, dark eyes, on a tennis court
Negative:
Just use photon's one as well. Nowaday models don't need a full novel quantity worth of words in prompts to gen correct pics.
cartoon, painting, illustration, (worst quality, low quality, normal quality:2)
I just added EasyNegative embedding on my everyday usage (plus more_details lora and the dynamic thresholding extention). I find results better fit my taste (I didn't use them for the illustrative pics though. I wanted them to be raw).
As usual for every checkpoint, use hires fix. And if you want awesome perfect faces from distance, use Adetailer (I don't, it's not working on my rig anymore. AMD powered):
https://github.com/Bing-su/adetailer
Of course if you like to write novels..... feel free. It's your usage, not mine. I don't mind what you do. Experiment, post and share, or not. :3
So... that's it I guess.
Hugs, or trout slaps.
