Full Training Tutorial and Guide and Research For a FLUX Style
Hugging Face repo with all full workflow, full research details, processes, conclusions, checkpoints,comparisons, prompts and more details > https://huggingface.co/MonsterMMORPG/3D-Cartoon-Style-FLUX
Trigger word : ohwx 3d render
Last image is training dataset as a grid
This is a training of a public LoRA style (4 separate training each on 4x A6000).
Experimenting captions vs non-captions. So we will see which yields best results for style training on FLUX.
Generated captions with multi-GPU batch Joycaption app.
I used my multi-GPU Joycaption APP (used 8x A6000 for ultra fast captioning)
https://www.patreon.com/posts/110613301

I used my Gradio batch caption editor to edit some words and add activation token as ohwx 3d render
https://www.patreon.com/posts/108992085

The no caption dataset uses only ohwx 3d render as caption
I am using my newest 4x_GPU_Rank_1_SLOW_Better_Quality.json on 4X A6000 GPU and train 500 epochs - 114 images
https://www.patreon.com/posts/110879657

All trainings are saved as Float and 128 LoRA Network Rank thus they are above 2GB per checkpoint
Inconsistent Dataset Training
This is the first training I made with the below dataset
Inconsistent-Training-Dataset-Images-Grid.jpg
When you pay attention to the grid image above shared, you will see that the dataset is not consistent
The training dataset with used captions (only for With Captions training) can be see in below directory
It has total 114 images
This training total step count was 500 * 114 / 4 (4x GPU - batch size 1) = 14250 steps
It took like 37 hours on 4x RTX A6000 GPU with slow config - faster config would take like half
There were 2 trainings made with this dataset. Epoch 500 checkpoints are named as below
SECourses_Style_Inconsistent_DATASET_NO_Captions.safetensors SECourses_Style_Inconsistent_DATASET_With_Captions.safetensors
Their checkpoints are saved in below folders
Training-Checkpoints-NO-Captions Training-Checkpoints-With-Captions
Its grid results are shared below
Inconsistent-Training-Dataset-Results-Grid-26100x23700px.jpg
When you pay attention to above image you will see that it has inconsistent results
Consistent Dataset Training
After I noticed that the initial training dataset was inconsistent i have pruned the dataset and made it much more consistent
Fixed-Consistent-Training-Dataset-Images-Grid.jpg
When you pay attention to the grid image above shared, you will see that is way more consistent, still not perfect though
Now it has total 66 images
The training dataset with used captions for this training (only for With Captions training) can be see in below directory
Fixed-Consistent-Training-Dataset
This training total step count was 500 * 66 / 4 (4x GPU - batch size 1) = 8250 steps
It took like 24 hours on 4x RTX A6000 GPU with slow config - faster config would take like half
There were 2 trainings made with this dataset. Epoch 500 checkpoints are named as below
SECourses_3D_Render_Style_Fixed_Dataset_NO_Captions.safetensors SECourses_3D_Render_Style_Fixed_Dataset_With_Captions.safetensors
Their checkpoints are saved in below folders
Training-Checkpoints-Fixed-DATASET-NO-Captions Training-Checkpoints-Fixed-DATASET-With-Captions
Its grid results are shared below - this one includes results from inconsistent dataset as well
Fixed-Consistent-Training-Dataset-Results-Grid-50700x15500px.jpg
When you pay attention to above image you will see now it is way more consistent
Best Checkpoint And Conclusion
When inconsistent dataset was used, training with captions yielded way better results.
However, when training made with a consistent dataset, no captions yielded better and more consistent results with early epochs.
Thus I concluded that, epoch 75 of no-captions dataset is best checkpoint
Here below comparison images for fixed dataset
Fixed-Consistent-Training-Dataset-No-Captions-Only-Grid.jpg
Fixed-Consistent-Training-Dataset-With-Captions-Only-Grid.jpg
Best checkpoint download link : Training-Checkpoints-Fixed-DATASET-NO-Captions/SECourses_3D_Render_Style_Fixed_Dataset_NO_Captions-000075.safetensors
75 checkpoints is equal to 75 * 66 / 4 = 1238 steps
Tutorials To Train Your Style
1 : https://youtu.be/bupRePUOA18
FLUX: The First Ever Open Source txt2img Model Truly Beats Midjourney & Others - FLUX is Awaited SD3
2 : https://youtu.be/nySGu12Y05k
FLUX LoRA Training Simplified: From Zero to Hero with Kohya SS GUI (8GB GPU, Windows) Tutorial Guide
3 : https://youtu.be/-uhL2nW7Ddw
Blazing Fast & Ultra Cheap FLUX LoRA Training on Massed Compute & RunPod Tutorial - No GPU Required!
The dataset can't be used commercially

Grid Testing Prompts - Example Images Picked From Grid - No Cherry Pick
a ohwx 3d rendering of a car
a car rendered in ohwx 3d style
a ohwx style car image
a ohwx render of a car
a ohwx car
a ohwx 3d rendering of a chest, depicted in a cartoon style. The background is a plain white, making the chest and its contents stand out clearly. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give the chest a realistic, three-dimensional appearance. The metal bands and rivets add a sense of realism and durability to the chest. The image is vibrant and eye-catching, inviting the viewer to imagine the treasure within. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold, with a focus on oranges, browns, and golds to create a sense of warmth and excitement. The overall mood is one of excitement and discovery.
a ohwx 3d rendering of an airplane, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a battleship, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a robot, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a dog, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a cat, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of an axe, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a house, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a dragon, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a flower, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a rose, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a tank, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a computer, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a graphics processing unit (gpu), depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a fork, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a lock, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
a ohwx 3d rendering of a umbrella, depicted in a cartoon style. The background is a plain white. The overall style is playful and whimsical, with clean lines and bright colors, suggesting a fantasy or adventure theme. The illustration is highly detailed, with a focus on textures and shading to give a realistic, three-dimensional appearance. The image is vibrant and eye-catching. The illustration is likely used in a digital context, such as a game or a children's book. The colors are bright and bold to create a sense of warmth and excitement.
Downloads last month
-
Downloads are not tracked for this model.How to track
Inference API
Unable to determine this model's library. Check th
Description
FAQ
Comments (20)
Pal... who the fuk is going to train for 37 hours ? Are you autistic ? You cant be real with this.This style does not need that many pics, its just 3d render.Some of your stuff is useful but this is insanity.
niccc please read the article without skipping a word and reply a comment again. you will understand purpose of this LoRA
Bro chill the flux out! The man is doing RESEARCH, he doesn't just watch YT videos and follows the pack, he tests shit on his own to find out what works.. People like him is the reason why a lot of us know what works and what doesn't.. He's simply sharing his results which he posts for free!
I regularly train large finetuned models that can take up to 80 hours on a single gpu. There are treatments available for ADHD.
@SouthbayJay “research” loul, he's just a wanker with a big graphics card looking to make a buck.
He's learned nothing more than the others.
Honestly. You, me and anyone can get this with just 32 dimensions -or even lower as 16 (same alpha)-, less number of images on dataset )if you want) and of course... what bothers me... do you really need to train this on MULTI GPU? On RTX A6000's? This is a mistake and a bad follow up for people starting to train. It's just an opinion though.
multi gpu totally depends on how much longer you want to wait, how many images you have. if you read the article you will notice that i picked 75 epoch as best. it would be 75 * 66 = 4950 steps which is totally doable on a single GPU
@SECourses Yeah! of course I readed it but you should point (as this is a some sort of a guide) that this is also doable on consumer GPU's (taking more time, yup) and not looking like you need to rent to train. I already know that as I also know that you don't need 128 dimensions or caption 3d render when the target style is the same style, but we shouldn't assume everyone should know this or know this. Take this as a friendly advice and not an attack U^^.
@LDWorksDavid well that is why watching tutorials is super important because i explain everything you just said in video :)
Amazing work thank you for all the info! Don't listen to the haters, the AI world needs people like yourself that put in the actual work, research every angle and thankfully you openly share the outcome. I appreciate you!!! Keep up the good work!!
thank you so much
Thanks for sharing your experiences, the style looks pretty amazing and I'll use it despite the huge size because of that. I'm curious to see if you can optimize this some more.
This is a great resource. Thank you a lot. 5K steps is pretty normal for full training a dataset. Very easily can be done in one night on Rtx 4090.
100% you can do easily in few hours. thank you so much for the comment
Thank you for your Wonderful work. Really happy to see your custom Lora and findings on Civit AI. Go for hundred more! The sky is the limit
thank you so much. yes i have bigger lora and fine tuning plans hopefully
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.

















