Photo-Realistic SDXL LoRA for Social Media
FameGrid (Photo Real) is a meticulously crafted SDXL LoRA, fine-tuned on a hand-picked dataset of 1,000 high-quality social media influencer images—a 3-month-long project fueled by precision and passion! 🔥
For the next level, check out the Ultra Real Version—available at [FameGrid UltraReal] —which is highly fine-tuned for even more realistic images, with better consistency and sharper poses.
Famegrid Checkpoint Available -https://civarchive.com/models/1693257/famegrid-sdxl-checkpoint
⚠️ Important Notes
Use Automatic1111 for exact results.
Some comfy users have reported issues when using a checkpoint with a built-in VAE. If you experience problems, try removing any external VAE if your checkpoint already includes one.
Low CFG values give more realism but may reduce clarity. Higher CFG improves clarity but can introduce artifacts and reduce realism.
Play around with the model weights to balance between generalization and the intended effect.
SFW WORKFLOWS
https://www.dropbox.com/scl/fi/g7ihe8yhzjd908tbukcxq/FameGrid-WorkFlows.zip?rlkey=7nxsqzhsv3jr7pz5fc2bxpjb9&dl=0
NSFW WORKFLOWS
https://www.dropbox.com/scl/fi/oedzv008omv3xzpr37jwc/68-FameGrid-Workflows.zip?rlkey=xhita5jz9i7985lf6dzt921hj&dl=0
Why FameGrid?
Every image in the base dataset was chosen with a sharp eye for quality, focusing on:
✔ Great poses 🕺
✔ Clean composition 📸
✔ Top-tier aesthetics ✨
This model generates photorealistic shots that feel right at home on the grid: natural poses, vivid style, and a polished look. It shines for:
📷 Amateur & professional-style photos
🛍️ General e-commerce product shots
🌐 Virtual photoshoots & AI influencers
✨ Creative projects & social media content
Want to craft the next big AI influencer or Instagram model? This is your go-to. 🚀
⚙️ Recommended Settings:
Weight: 0.2-0.8
CFG Scale: 2-7 (adjust for realism vs clarity)
Sampler: DPM++ 3M SDE
Scheduler: Karras
Trigger:
"IGMODEL"
FameGrid (Photo Real) is built for creating realistic, standout visuals—simple as that. 💡
Description
FAQ
Comments (8)
Am I outdated, cause I do not remember ai being able to generate faces that are in weird angles
Why do you use Pony prompting for SDXL? Was this LoRA trained that way, and is it better to type all these scores?
When using lustify or big lust checkpoints, you can be because I believe they are merges from a pony model
Or no, that's wrong. Big Lust is a merge of big asp and lustify. Big Asp was trained on those style tags, see
https://civitai.com/models/502468/bigasp
Quick review for the creator of FameGrid and FameGrid UltraReal
First i installedc the free version and used with the "workflows" provided in the archive.
I must say, i was expecting more from somebody who also is selling online a service.
I have a online business too but sometimes you just need to put some more effort into
whatever you provide, even if it's a free service, not to mention a paid one.
So what are the negative points :
- visually, the workflows are a complete mess. nodes are scattered here and there all over the place
-missing nodes that could increase the workflow functions ( save image),
-some nodes are broken ( had to replace with working ones), for example image film grain nodes and might be more but i didn't had time to test them all
Positive points : Lora generates great images!
I thought that maybe purchasing the Ultra/Pro whatever it is called version i will see a change.
Well..not that much.
After purchasing i got a download link with the lora..and that's all.
Some would say that for a $30 product you could receive more.For example a clean
good looking workflow. A readme file ( because not everyone is an expert in comfyui. i don't say i am
a beginner or that i need it, it's just as a business approach and work ethics)
The lora itself : it doesen't produce such a big difference ( as advertised), i tested them both with same settings
i could not notice a big difference.
Images that lora has been trained are mostly from Onlyfans accounts, which migth be an issue at some point,
depending on how they are used
As a conclusion, the lora performs well, and that's all. For the rest of the service...really disappointed!
If you wanna see ho it's done, and what work ethics is and how a product is sold ( or giving it for free ) maybe you should
check Stable Yogy's page. There's a reason for why he is on the top creators on Civitai.
I admit I'm no expert in ComfyUI. The workflows I created for ComfyUI were in response to requests from users and weren't a part of the UltraReal download package. It’s crucial to note that all the images showcased on our page were generated using Automatic 1111, which operates on a different architecture from ComfyUI. This difference can significantly impact the output, and while ComfyUI can yield great results, it functions differently.
I understand your points about the visual mess in the workflows and the missing or broken nodes. That’s definitely on me to fix, and I'm currently working on refining these workflows based on user feedback like yours. Also, the inclusion of more intuitive elements like a save image node and a more comprehensive readme file sounds like a wise enhancement, especially for users who aren't as familiar with ComfyUI.
That being said, all feedback is much appreciated.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.