Mystic WF - Full flux workflow版本v1.0 (ID: 991542)

V2: The workflow has been slightly expanded to include additional features. I have integrated ControlNet's Canny and Depth modules, inpainting capabilities, the Florence prompt generator, and a signature feature at the end. Despite these enhancements, the workflow remains straightforward and user-friendly, with all processes fully automated.

---

Full flux workflow, with upscaler and saving metadata for civitai. It will clean the lora lines that sometimes happen on Flux, can upscale quit a bit and can do img2img.

This workflow is a merge from my old workflow and this beautiful one https://civitai.com/models/872281?modelVersionId=981799 made by https://civitai.com/user/Catz . Please guys donate some buzz to Catz he is a monster.

In the picture I put some numbered marks:

1. ControlNet - Deph

2. ControlNet - Canny

3. Inpainting using Redux

4. Img2Img area, use if you want or just leave on by pass.

5. By pass area, you can select what to run on each time. Usually I activate only 1 until I got a good seed and then activate all the others

6. Your prompt

7. Save file and metadata for civitai. You can configure what to save. You only need activate one save. Or leave all the yellow boxes active and save at each step

8. First upscale, this one will increase the details by a good amount. Change the denoise settings as you like, usually is better to leave at 0.52~0.62 but this may change for each style

9. This step can clean the lora lines if you got no luck on your generation. You may need to change the denoise here too. Usually I leave it on 0.1~0.25.

10. Extra upscale, sharpness and color adjustments,

11. Add your signature or watermark on your pictures

12. Save file and metadata for civitai. You can configure what to save. You only need activate one save. Or leave all the yellow boxes active and save at each step

13. Load an image and start from 9

14. Load an image and start from 11

15. Load an image just to save the civitai info

16. Florence prompt generator from an image

17. Your loras, there are 4 but you can increase and just put more groups to the left. Be careful, to many loras may produce some lines on your picture. It happens in flux all the time.

All steps may be by passed and this works super nice.

I added all files on the v1 zip file for convenience, the v2 has only the workflow.

1. Put the luts inside your ComfyUI/models/luts directory, if you don't have one just create it.

2. The upscale models go inside your ComfyUI/models/upscale_models

3. For a grain image you must experiment with some pictures and see what you think is best for you. Some people just don't like it and remove it, some people love it. I think that each style and picture may be a little different in this regard.

描述:

训练词语: