theMickey_
Engaged Member
- Mar 19, 2020
- 2,340
- 3,163
- 357
I've been messing with Flux in WebForge and I'm really happy with the results, and the growth in available checkpoints and lora for making it more suitable for NSF work. I don't want to go and learn Comfy UI in addition to all I'm teaching myself already to get my VN going.
I've been waiting patiently for ControlNet to be up and working for Flux, and now Black Labs go and drop these great new checkpoint backed in control tools. But I can't see any way to use them in Forge. Can any of the experts here help? I've been reading this thread off and on for ages, maybe this has come up and I've missed it so a link back would be a great response if its been covered already thanks and sorry.
I think I'm going to have to learn Comfy
Any good guides out there?
I've been waiting patiently for ControlNet to be up and working for Flux, and now Black Labs go and drop these great new checkpoint backed in control tools. But I can't see any way to use them in Forge. Can any of the experts here help? I've been reading this thread off and on for ages, maybe this has come up and I've missed it so a link back would be a great response if its been covered already thanks and sorry.
I think I'm going to have to learn Comfy
Any good guides out there?
- Sep 20, 2021
- 908
- 4,389
- 639
I just wanted to mention something perhaps non-obvious. My uploads are 'open source'.
You can download almost any of the PNG's I've posted here on F95, drag them into "PNG info" in Forge and instantly see all of my prompt, loras, lora strengths etc.
Click 'send to txt2img' and you can even re-generate an identical render locally, perhaps tweaking the prompt or Lora strengths to see the effects. All you need to do is ensure you're using the same checkpoint (usually Flux 1 Dev FP8 for me) and Loras (usually found on Civitai).
If you find something out of place or odd in the text in my prompts, it's probably because experimentation showed me that this combination of tokens helped to guide the 'thought journey' of the diffuser in just the way I wanted. Or it was a mistake and fixing it would have de-stabilized the render results more than it was worth. 8)
Lora trainers use a wide variety of software, tools and techniques and their results can vary hugely, with some really idiosyncratic behavior in some cases.
Generally, well structured plain English prompts do work really, really well in Flux. But season them with some seemingly surrealist tweaks and weird, sometimes wonderful things, can happen.
You can download almost any of the PNG's I've posted here on F95, drag them into "PNG info" in Forge and instantly see all of my prompt, loras, lora strengths etc.
Click 'send to txt2img' and you can even re-generate an identical render locally, perhaps tweaking the prompt or Lora strengths to see the effects. All you need to do is ensure you're using the same checkpoint (usually Flux 1 Dev FP8 for me) and Loras (usually found on Civitai).
If you find something out of place or odd in the text in my prompts, it's probably because experimentation showed me that this combination of tokens helped to guide the 'thought journey' of the diffuser in just the way I wanted. Or it was a mistake and fixing it would have de-stabilized the render results more than it was worth. 8)
Lora trainers use a wide variety of software, tools and techniques and their results can vary hugely, with some really idiosyncratic behavior in some cases.
Generally, well structured plain English prompts do work really, really well in Flux. But season them with some seemingly surrealist tweaks and weird, sometimes wonderful things, can happen.
How do i make a smaller futa girl fuck a bigger girl?I put size difference in the prompt,but its always the bigger fucking the smaller...
DD3DD
Active Member
- Apr 23, 2019
- 949
- 1,795
- 286
Aaah... It always comes down to the classics no matter what... Little Spoon or Big Spoon?
what are yall using Flux models in? a million Flux models getting uploaded to civitai. but i have yet to get them to work. not to mention they're gigantic. tried in RuinedFooocus, SwarmUI, and nothing but headaches and errors and gotta do 50 things to set it up(yes im exaggerating). and not a COmfy user so thats out. don't understand why A1111 ddint add it to their version...
- Sep 20, 2021
- 908
- 4,389
- 639
yes tried in forge as well. still issues. do you mind sharing ur Forge setup? so i i can get the idea
this thread mentions what models work with ruinedfooocus. but i couldn't get any to work.
You must be registered to see the links
- Sep 20, 2021
- 908
- 4,389
- 639
Defaults, really.
Install Forge, drop the required F1D, vae, and clip_l checkpoints in place, type a prompt, click generate and voila.
When you select 'Flux' mode (radio buttons at the top) in recent versions of Forge, I think it automatically sets the schedule to simple and sampler to Euler.
I do tweak steps, and occasionally Dynamic CFG, but generally the only thing I change is tokens and lora strengths.
I rarely vary checkpoints (reading instructions and changing settings appropriately when I do) and most Flux loras work, to varying degrees, out of the box.
Good luck!
Install Forge, drop the required F1D, vae, and clip_l checkpoints in place, type a prompt, click generate and voila.
When you select 'Flux' mode (radio buttons at the top) in recent versions of Forge, I think it automatically sets the schedule to simple and sampler to Euler.
I do tweak steps, and occasionally Dynamic CFG, but generally the only thing I change is tokens and lora strengths.
I rarely vary checkpoints (reading instructions and changing settings appropriately when I do) and most Flux loras work, to varying degrees, out of the box.
Good luck!
I am new to AI, I want to ask some question:
1/ Can you guys show me the necessary VGA, CPU and RAM to run AI?
2/Can AI build porn video scenes?
3/Can AI convert real video to 3d video?
1/ Can you guys show me the necessary VGA, CPU and RAM to run AI?
2/Can AI build porn video scenes?
3/Can AI convert real video to 3d video?
- Sep 20, 2021
- 908
- 4,389
- 639
1) RTX 3xxx upwards, priority is size of video RAM. Fast CPU, lots of system RAM. Just look at the requirements for the model/tool just as you would with any other software.
2) Kinda. I guess. Not really.
3) Do you mean SBS for VR or something? Don't think we're there yet.
I've been using webforge, but I think I may need to use Comfy, some of the new Flux Tools don't seem to have a way to work in Forge. I'm not Comfy user either and I don't want to be. But I think I may have to.
devilkkw
Member
- Mar 17, 2021
- 340
- 1,151
- 258
I had same things time ago, switched to a1111 to cui was a bit pain in frist time, but when you understand how powerful is CUI, you don't turn back again.
The option for making image change your world.
Start with simple workflow and compose it like you need.
Also with cui I'm able to run flux, and SD 3.5 on my 6gb vram.
Just start simple and understand how it works. And for help just search and ask here.
- Jul 4, 2017
- 3,350
- 6,437
- 707
You could try asking in a forum of people who may have downloaded such models and could possibly still have them saved in storage. But it would be wise to clearly specify what model you are looking for.
Alternatively, perhaps there is some site which worked to actively mirror hugging face content... I don't know of any however.
Understood, I am in search for this model :
Its name was AngleDroit - V1
You must be registered to see the links
Its name was AngleDroit - V1