>>> If you think my content is worth it, consider donating to my Patreon or Ko-Fi to help fund the training of new LoRA's or porting existing LoRA's over to other base models! <<<
Trigger Phrase (must be included in the prompt for the full effect)
v13: casual snapshot
v12: casual snapshot
v11: snapshot photo
Recommended inference settings/workflow
use the default "9B Text-to-Image" ComfyUI workflow (found here) with an added "Load LoRa" node and the following settings (which are the default settings for the model as per its HuggingFace page):
euler (sampler), 50 steps, cfg 4, 1 megapixel resolution (i.e. 1024x1024 for a 1:1 aspect ratio)
How to prompt:
Ask your LLM of choice to generate you a "literal, technical" prompt that is about 2 paragraphs long of which each paragraph has about 4 sentences that are written in a natural language style.
Latest version notes and model page updates:
major changes: completely revamped entire training workflow, everything is better: quality, coherence, style likeness, etc...
added a "How to prompt" section
LoRa description
This LoRa model was designed to generate images that look like casual snapshot photos taken with a Samsung Galaxy A52 smartphone.
Disclaimer:
I do not claim to be the current owner, original creator of, person or legal representative of a person depicted in, any of the concepts that my models, including this one, aim to emulate nor the training data used to train this model. I do not claim that any of my models have received an endorsement by those respective individuals. I also do not claim that the emulation attempted by my models represents a 100% accurate depiction of the original concept in question nor do I claim that the quality of my emulation reaches the same quality of the original concept. All credit goes to the respective current owners, original creators, or people and I encourage you to support them in any way you can.
If you are the current owner, original creator of, person or legal representative of a person depicted in, any of the concepts that any of my models aims to emulate and would like that specific model to be removed, then please leave me a private message here or on Reddit with proof of authenticity of your identity and claim and I will remove it.
Additionally, I do not endorse my models being used to violate the law or enact still legal, but immoral acts such as the spread of misinformation using deep fakes.
Description
FAQ
Comments (22)
Looks great, thanks for this!
Nice work, but 50 steps?? Jesus is this any better then Qwen + loras at 4 steps?
Use cacheDIT : https://github.com/Jasonzzt/ComfyUI-CacheDiT
i'm using the node with zimage setting, it can speedup your gens by up to 1.4x-1.7x with minimal quality lost. It cuts down my gen time from 90 sec to 33 sec for 50 steps klein 9b.
Or you could drop cfg 4 3 or 1 and do less steps cfg1 is 4 ateps
You can reduce the number of steps you know. Those are just the recimmended default optimum...
Hello mate!
I've been waiting for you to get into the Klein game! How is it going so far?
Any secrets so far for workflow or training?
Klein is so far imho the best model I have ever trained. By far the easiest and doesnt require many resources either.
@AI_Characters Good to know, thanks. I tried to make a slider... It went EXTREME in a couple steps. I can't yet figure it out how to slow it down. I think the issue is Differential Guidance. Not a lot of data on it yet.
cant wait for the qwen verison!
srry for the dumbass question, but does it work on the 4 steps distilled version? or we need to wait for that?
I dont know I havent tested it.
@AI_Characters It works, works well on the distilled 9B
Will this work with Klein 9B distill too?
You pretty much should only use distill for generation, base should be for training.
@brnfd24434343d Thanks
I'm using V9.0
Works fine with Klien 9b distilled, I use 6 steps though.
Thank you.
I like that you re-used the samples on the different versions. For my money... This looks better than the new Qwen 2512 version, and that looked really good.
No I agree the Qwen-Image-2512 version doesnt look as good but I blame it on the model not my training. Klein is just simply better.
can this be used as img2img and not juist text2img? sorry if it's stupid but I am new. basically I have a character photo and I want to create more photos of her taht look real
exactly what im asking too, i want to to turn some of my image into amateur one
It can definitely be used for Klein image editing. I use it all the time!
Details
Files
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
Mirrors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
KLEIN_Smartphone.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
smartphone-snapshot.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_STYLE_SmartphoneSnapshotPhotoReality_v9_TRIGGER$amtr snapshot photo$.safetensors
KLEIN_Smartphone.safetensors
smartphone_snapshot_v9.safetensors
FLUX.2-klein-SmartphoneSnapshot.safetensors













