Furkan Gözükara
patreon
62 Prompts tested on all experiments (fully public - open access - visit OLDEST COMMENT - all raw Grids shared) to find best Sampler + Scheduler for Stable Diffusion 3.5 Large - SD 3.5 Large FP16 vs Scaled FP8 compared - T5 XXL FP8 vs Scaled FP8 vs FP16 compared - FLUX FP16 vs Scaled FP8 compared
🕑 Added 2024-10-26 12:45:00 +0000 UTC62 Prompts tested on all experiments to find best Sampler + Scheduler for Stable Diffusion 3.5 Large and SD 3.5 Large vs FLUX DEV > https://youtu.be/-zOKhoO9a5s
FLUX Dev vs SD 3.5 Large fully compared.
SD 3.5 Large FP16 vs Scaled FP8 fully compared.
T5 XXL FP8 vs Scaled FP8 vs FP16 fully compared.
FLUX FP16 vs Scaled FP8 fully compared.
Also how to install SwarmUI on Windows, Massed Compute and RunPod shown in the tutorial.
I have shown how to use FLUX and SD 3.5 Large in details as well.
Stability AI published their most power newest model Stable Diffusion 3.5 Large. This model unlike FLUX is full model not distilled and has huge potential. I have done extensive research and publishing all of it in this video regarding how to use SD 3.5 Large with the best settings. Moreover, I am sharing how to use FLUX DEV with the best possible configuration as well. Moreover, I am making a huge comparison between SD 3.5 and FLUX and you are going to learn who is the winner.
🔗 Full Open Access Public Article Used In The Video ⤵️
▶️ https://www.patreon.com/posts/106135985
The biggest advantageous of SD 3.5 Large are its license and also it is non-distilled thus fully trainable. So I am expecting amazing community Fine-Tuned models. Hopefully I will also fully research SD 3.5 Fine-Tuning / DreamBooth & LoRA trainings workflows and make videos and compare with FLUX.
I will show how to use Stable Diffusion 3.5 Large and FLUX via the SwarmUI. SwarmUI is the most advanced, most optimized UI with most amazing and easy to use interface. It uses ComfyUI back-end thus supports all of the newest generative AI models with amazing speed and so easy way to use.
🔗 Full Open Access Public Article Used In The Video ⤵️
▶️ https://www.patreon.com/posts/106135985
🔗 Unified Models 1-Click Robust Auto Downloader Scripts ⤵️
▶️ https://www.patreon.com/posts/114517862
🔗 Full FLUX Tutorial — FLUX Beats Midjourney for Real ⤵️
▶️ https://youtu.be/bupRePUOA18
🔗 Main Windows SwarmUI Tutorial (Watch To Learn How to Use) ⤵️
▶️ https://youtu.be/HKX8_F1Er_w
🔗 Cloud SwarmUI Tutorial (Massed Compute — RunPod — Kaggle) ⤵️
▶️ https://youtu.be/XFUZof6Skkw
🔗 SECourses Discord Channel to Get Full Support ⤵️
▶️ https://discord.com/servers/software-engineering-courses-secourses-772774097734074388
🔗 SECourses Reddit ⤵️
▶️ https://www.reddit.com/r/SECourses/
🔗 SECourses GitHub ⤵️
▶️ https://github.com/FurkanGozukara/Stable-Diffusion
Video Chapters
0:00 Introduction to the SD 3.5 vs FLUX majestic tutorial
1:54 Windows installation of SwarmUI and the main part starts
4:08 How to download Stable Diffusion 3.5 Large and FLUX FP8 Scaled Version
5:05 Which model variants are recommended for which GPU and why
6:45 How to update and start SwarmUI and start using the models
7:27 How did I find the very best Stable Diffusion 3.5 Large model configuration — grids publicly shared
7:54 Which configuration are best for SD 3.5 Large
9:29 How to see details of what is happening via server — logs — debug
10:22 How to use Stable Diffusion 3.5 Large LoRAs and FLUX LoRAs or any LoRA model
10:33 Where LoRA files are need to be put in SwarmUI
10:43 How to set LoRA and LoRA scale / strength / power / impact
11:41 How to use FLUX DEV and variants with the best configuration
13:58 Be careful of selecting only compatible LoRA models
14:31 Detailed Grid comparisons start
16:12 Detailed comparison of Stable Diffusion 3.5 Large vs FLUX DEV
19:31 How FLUX DEV becomes way more realistic and better looking after Fine-Tuning / DreamBooth training in a short time
21:30 How to install and use SwarmUI with Stable Diffusion 3.5 Large and FLUX on Massed
Compute
26:34 How to start SwarmUI on CloudFlared Link to use it in your Computer browser or in your Table or Phone and such
27:45 How to use LoRAs on Massed Compute like SD 3.5 LoRA or FLUX LoRA
28:54 How to install and use SwarmUI with Stable Diffusion 3.5 Large and FLUX on RunPod
Comments
Furkan Gözükara
nice but not published yet : Provide a script for inference on the user's own video
Philip Heggie
Hi I'd love you to do this video enhancer but output mp4 https://github.com/DachunKai/EvTexture
Furkan Gözükara
we cant train fp8 all trainings are done at bf16 kohya doesnt know how to do fp8 training :d
Mario Santiago
Do you have comparison of flux finetune done in fp8 vs fp16?
Furkan Gözükara
true. as i said in video we can expect good SD 3.5 Large fine tuned checkpoints
cool1
If SD3.5 large is being compared to Flux, isn't one thing that could be compared the license, eg. if you wanted to use one of them commercially then isn't SD3.5 Large free to use commercially (if you earn less than $1 mill a year) but you can't use Flux Dev commercially (or at least not for free). Though I think if you want to use SD3 Large commercially you have to register with them so it's a bit restrictive in having to do that. Though the Flux Schnell can be used commercially, so maybe SD3.5 Large could be compared to that too. And if SD3.5 can do better styles that's also good and the negative prompt option that Flux doesn't have. Though if Flux can do more accurate hands that's one advantage for that. It will be interesting to see how well SD3.5 Large can be trained for Loras/Dreambooth. I assume the current Kohya hasn't been updated for it yet (it has an SD3 option though). If SD 3.5 is a bit lower size, maybe it will be easer/faster to train (or easier at bit higher res. Though I think it's intended for only about 1K wide even though flux can output higher).
