Here's a question for you folks reading out there; how long have you been doing AI image gen for?
I've been doing it for 9 months (basically a lifetime ago!) and even now, thousands of image gens later, I'm STILL learning, and as it happens, forgetting stuff.
With the sheer amount of material being uploaded to CivitAI each day, it's easy to miss something awesome, especially a checkpoint. Everyone has their own faves, but there are so many great merges out there that with only a handful of downloads.
Everyone knows and loves epiCPhotoGasm, Realistic Vision, MeinaMix, RealCartoon3D and PicX_real; but where's the love for other checkpoints that never got that popular?
So this series (I hope) will be a look back at just that; the Under-appreciated, Forgotten, and Overlooked (UFO) checkpoints. Especially ones contributed by users no longer active on this platform, so maybe they finally get the credit they deserve. Hopefully this introduces a new fave to you, or reminds you of one you've forgotten.
Lemme know in the comments if you did try em (and liked/hated it), or if you have any suggestions for future reviews. Bookmark/save this article if you'd like to get notified of future updates!
Oh, and please like/follow/subscribe/engage with your fave creators on here, it might just make their day.
#1: Sweet Mint by Notstable (SD 1.5)
The first of 2 checkpoint merges from user Notstable (the other being the less refined Sour Mint), uploaded Sep 29, 2023, with less than 1,300 downloads to date. No details were provided about what went into the merge, but the aim was to generate images that provide an "impression of an artistic drawing that came to live without making it look too cartoony or too realistic".
Why do I like it? : It has nice details, and has a slightly different aesthetic with visually appealing subjects. Does really well with female subjects, not so much for male ones.
Why does it deserve credit?: It's a pretty great merge for someone who only had 2 uploads and 4 posts in total (and doesn't seem to have been actively uploading for the last 6+ months).
Random sample images included for purposes of this review.
#2: MoltenArtv2 by MoltenHeart (SD 1.5)
The second version of the already great MoltenArt v1 (from user MoltenHeart), uploaded Jan 25, 2024 with less than 550 downloads to date. No details were provided about what went into the merge, but the result was a merge that "..can handle fantasy, Loras, lycoris, textual inversions, Booru tags, simple or complex prompts, etc".
Why do I like it? : Just like the author intended, it works really well for prompts aiming for the fantastical, with great backgrounds and lighting capability (also does very well with female subjects)
Why does it deserve credit?: It's a great merge that responds well (even to newbie prompts) and is the only upload on the site from MoltenHeart (so it's kinda a collector's item)
Random sample images included for purposes of this review.
#3: fennfoto by Fenn (SD 1.5)
Similarly, the second version of the first-gen fennfoto v1 from user Fenn), was uploaded Dec 5, 2023 has just over 1760 downloads to date. The author describes it best, so I'll quote him/her here; "I've looked for every piece of garbage unique dataset I could find.
On here, HuggingFace, etc. and merged them endlessly until I arrived here. If someone else said it was trash and the outputs were horrible, I merged it. There's easily hundreds of datasets involved in the creation of this, each bringing a unique element into the fold."
Why do I like it? : It's a personal fave that works really well for photorealistic images, with great detail and depth. Plays well with LoRAs (including my own Shot by Elsa Bleda photography LoRA)
Why does it deserve credit?: There is definitely a sense of intention around this merge, as the author only used original trained datasets and worked on refining the merge for a year (with great results)
Random sample images included for purposes of this review.
