Been watching this lovely lady since the unreal engine tutorials. I’m all in with AI now and this is precisely the video I needed. Thanks for what you do!
@@promptmuse Definitely. I'm new to game design but I suck at digital drawing and painting so this tool is going to help me visualize the game concept art I have in my head. Playing around with it right now.
Thank you for all the hard work and curating/testing, excellent video, kudos. I can finally feel safe to experiment and achieve results as a frugal artist in this rapidly changing AI space. I encourage you and cheer you on. Have a nice one! :>
Thank you so much what a lovely comment to wake up to! I try my best even though my time is limited 🤍 so it’s really nice to hear - thank you. I also hope this proves useful for you in your endeavours.
Thank you for showing me how to do all this. Your tutorials are amazing and I’ve learned so much from you. Thank you 🙏 Ai has been so fun and it was your channel that made it accessible to my head to wrap around. I hope you the very best and that life goes superbly for you because you deserve it.
@@promptmuse hey! I don't know if you can help me. But I just installed Forge, run it and get an "access denied. couldn't launch python" message. Do I need to have Python installed to be able to use Forge?
Forge is great for flux. I can run fp8 on my 3060ti 8gb vram, since it has memory swapping. But if I’m strapped for system resources the Q4 GGUF version works well too.
@@MonzonMedia I haven’t actually gone back to comfy for awhile…….. I do like comfy for being able to control more but the ease-of-use with Forge and being able to manage memory so easily is +
@@promptmuse I have a love hate relationship with comfy 😁 Forge has been my go to for awhile now. BTW they will support controlnet for Flux soon. You should also try Invoke AI as well. They have an amazing canvas. Great tool for more creative work.
2 месяца назад
Je tenais à vous remercier pour cette explication si lumineuse et ces liens très pertinents.🌹
@promptmuse this is awesome. But i have to wait a month for my new pc set up. Where is the Mac version of this? Can you point me towards the source? Thanks
So I got things working and tried using the retro anime lora, and the image comes up black. I copied and pasted from one of the examples, and clicked on the lora to add it, so I'm not sure what I did wrong. Any ideas?
This is interesting! Just make sure you download the LoRa correctly and that it’s got .saftetensor at the end of the file name. If you’re not logged in and you can right click on the download button, it will save a weird file. That’s not the Lora. Let me know how it goes , and if indeed this was the issue
Have you tried using Krita AI Diffusion as a graphical interface for SD or Flux? Having experience with graphic design and Photoshop, for me it's the best way to work with generative AI models. I find Forge UI a bit all over the place and not really practical.
Krita is fantastic! To be honest, I would like to do a video or perhaps something a quick comparison between all of the UIs as there are so many now, foocus / krita / comfyui/ invoke/ a1111 etc etc I think each one serves their purposes well. I agree, forge interface takes some time to get used to. Transferring over from automatic 1111 it does feel like coming home (without the bugs and smoother performance ). And then for animation that’s a bit more complex than this video. I would jump into comfyUI ( so I don’t have to pay for a service). They’re like Pokémon gotta catch them all
hmmm....everything was going great until sanity check....the model showed int he field...but after hitting generate it disappeared. command window says: Model type not recognized. checked the folder and the model name was now appended with the word "corrupt" at the end...giving it a third try
I Don't know if a "my computor problem" or a "file problem" but unpacking this file takes ALOT of time, it planed to take around 90 hours for me, never ever seen that before 😯
@@geekpotion 90 hours …. No, it should not take that long. Are you zipping? The folder? Is that what’s taking a long time? Should take a few minutes top. Do you have enough space on your hard drive?
Great explainer video ❤. Do I need stable diffusion installed on my system to be able to install Flux? Also, How to train a Lora on my own images on flux. Could you please also make A tutorial on that.?
@@promptmuse OK.. Thanks. I see some tutorials where they have mentioned installing the latest version of python for installing flux. Actually, I have been looking for the easiest way of installing flux, and there are so many tutorials out there with different ways of doing it, that it gets pretty confusing for a newbie like me. 😅
Oh yes, totally understand it can be confusing. If you install for Forge UI first that actually has the correct python version packaged in with it so you don’t need to do anything else 👍
Recommended settings if I'm running this setup with an RTX 3070? First generation I tried to do after installing the UI and the models was showing an hour plus to create results...
And at the end all I had was a black picture of the size specified. I'm assuming that I really need to get a card with more memory or an even more optimized model?
@@TheGodExperiments It's running fine for me on a 3060 Ti, so I would assume it's something with your settings or installation. Did you change the UI to Flux with the radio button? The sanity check image with her settings took around 2 minutes I'd say. I used the nf4 version.
I like your channel and maybe you like what we do. I myself have been working with AI for almost 3 years, but also with blockchain technology and how it can be used in the film industry.
I've just not had good results, perhaps it 'PINIC'. But for the next video will look into it more, as my subscription accidentally rolled over again for the month (Mybad) ha so I'll play with it a bit more!
@@SolearGnG the devs made it pay what you want so people can use it for free she's bringing in more people to try pureref and anyone who likes it can donate whatever they like - wouldn't it be more dishonest by saying its not free
@@SolearGnG It's not free for businesses, but it is free for personal use. That selected amount is just a donate button. When you enter 0 the total amount will literally say "free." lol
Do you think this will work on 2015 gtx980 8GB ? These only use GPU, rlght? I have an old laptop with i7 2.9 GHZ , GTX980 8 GB GPU and 32 GB RAM. But these do not matter, only GPU, am I right?
Give it ago, I cannot see why not. If you do go for it, just make sure you down the Flux model Nf4(as pointed out in the video). Make sure you got enough airflow for the laptop as it may get a lil warm :)
@@FunniestCatsandPets good luck, yes I would love to know if you works on your specs. If you do get any memory issues, you can decrease the weights in the UI. Also makes smaller images and then go about upscaling them after if you run into issues. Do not run any programs in the background either ( I’m sure you know this all already) 🤣 Looking forward to hearing back from you !
Quite possibly the most beautiful nerd that ever lived. Makes me want to lift weights and purchase minoxidil 😂 Games Master's 'Games Mistress' with a hint of John Bentley.
it seems when i try to run the sanity check it errors and when i check the checkpoint files i saved in stable-diffusion flux1-dev-bnb-nf4-v2.safetensors Full flux-dev checkpoint with main model in NF4.
@@blenderisfunty hey, try to re-download the checkpoint file again as sometimes it can become corrupt. It takes probably 45 minutes to download in total, sometimes longer depending on spec. Also, if there is a Wi-Fi interruption that can corrupt the file too 🤞 good luck let me know if this is a fix for you
@@promptmuse Iin firefox i used save link as and it didnt download anything so i clicked into the hugging face and downloaded from there to my stable-diffusion folder, now no files are corrupted but i i get an error Connection errored out. pop up toast on the webpage looks like this is where the command prompt gets hungup: [Unload] Trying to free 7725.00 MB for cuda:0 with 0 models keep loaded ... Done. [Memory Management] Target: JointTextEncoder, Free GPU: 3290.45 MB, Model Require: 5154.62 MB, Previously Loaded: 0.00 MB, Inference Require: 1024.00 MB, Remaining: -2888.17 MB, CPU Swap Loaded (blocked method): 3486.38 MB, GPU Loaded: 1740.62 MB Moving model(s) has taken 2.85 seconds Distilled CFG Scale: 3.5 Press any key to continue . . .
I’m still troubleshooting some speed issues since it’s currently taking me about three minutes per generation. However, I found that downloading the Flux file directly from lllyasviel (Hugging Face)-instead of using the "copy link" steps-helped me avoid the "corrupted file" issue entirely.
Hey, which model did you download? Sometimes if the model didn't completely download correctly (for whatever reason) and the file is corrupted, this can be an error. This could be the issue, if you retry the download again and let me know if the issue persists.
@@promptmuse I think that's exactly what went wrong. I guess the save link option doesn't work quite the same in Opera. I see now that the models downloading are much larger. My oversight. Thank you! I love learning from you!
@@agente_ai LOL I actually just allowed it to go to the hugging face link and clicked "download" and it's working great now! Opera is amazing but has its quirks for sure! Here's to FLUXing! Woot!
Sorry, great video but didn't work for me. Couldnt find the "sanity checks". Step number three just has a link to PUREREF and CIVIT AI. It wasnt even following my prompt, A prompt for an astronaught was turning out a grainy black and white portrait of a woman. Going back to SDXL. at least that worked.
Hey sorry to hear this, was the UI switch over to flux? If you still want to try, it might be worth trying to download the flux model again. The file can become corrupt if wifi is interrupted or incomplete download. You can use your SDXL in forge as well and switch back and forth.
Hey, just make sure that the file is probably downloaded, it does take some time as it quite big. Also, sometimes the file becomes corrupted if let’s say the Wi-Fi is interrupted and will give this error.
I've tried this several times and it just gave a 39kb file. click on the link to go to the page then where you see "This file is stored with Git LFS . It is too big to display, but you can still download it." click on the download link to start downloading the 12GB file.
Welcome bcak hun, loong time, and and nice to you come back again with a new tutorial
Been watching this lovely lady since the unreal engine tutorials. I’m all in with AI now and this is precisely the video I needed. Thanks for what you do!
Thank you, and I appreciate you watching from the start ! 3D and AI are so much fun, and I’m glad we can be a part of each other’s journey 🤍🙌
I just wanted to thank you for such an enlightening explanation and these very relevant links.
You are most welcome 🤍🙏
….as always , great clarity of installation setup walk through , and useful tips …….
Brilliant tutorial. Just what I was looking for. 👍
@@RealOGfikey amazing glad we found each other!! 🙌🙌🙌
@@promptmuse Definitely. I'm new to game design but I suck at digital drawing and painting so this tool is going to help me visualize the game concept art I have in my head. Playing around with it right now.
Fricking awesome! Thanks once again 👏
Thanks Ninjaaaaa !!!
@@promptmuse 😁👏👌
spanish: Eres la mejor explicando y he aprendido un montón de cosas, gracias por hacer lo que haces! i love you
Thank you for all the hard work and curating/testing, excellent video, kudos. I can finally feel safe to experiment and achieve results as a frugal artist in this rapidly changing AI space. I encourage you and cheer you on. Have a nice one! :>
Thank you so much what a lovely comment to wake up to! I try my best even though my time is limited 🤍 so it’s really nice to hear - thank you. I also hope this proves useful for you in your endeavours.
Thanks for a great guide! Well explained and with relevant links, and good tips on prompting, inpainting, Lora's, and animation 👍😊
Thank you for showing me how to do all this. Your tutorials are amazing and I’ve learned so much from you. Thank you 🙏
Ai has been so fun and it was your channel that made it accessible to my head to wrap around. I hope you the very best and that life goes superbly for you because you deserve it.
Amazing video! Really really helpful. Thanks a LOT!
Subscribed!
Love it , welcome aboard 🔥🙏
@@promptmuse hey! I don't know if you can help me. But I just installed Forge, run it and get an "access denied. couldn't launch python" message. Do I need to have Python installed to be able to use Forge?
@@leonardofavaretto2289 For me it just ran python automatically. Did you run the "update.bat" before the "run.bat" file?
This is incredible. A great tutorial demonstrating great usage. Is it possible to swap the image prompt data by using other images as data points?
awesome job, lifesaver really !!!!
Cheer John, thanks so much!
Forge is great for flux. I can run fp8 on my 3060ti 8gb vram, since it has memory swapping. But if I’m strapped for system resources the Q4 GGUF version works well too.
@@MonzonMedia I haven’t actually gone back to comfy for awhile……..
I do like comfy for being able to control more but the ease-of-use with Forge and being able to manage memory so easily is +
@@promptmuse I have a love hate relationship with comfy 😁 Forge has been my go to for awhile now. BTW they will support controlnet for Flux soon. You should also try Invoke AI as well. They have an amazing canvas. Great tool for more creative work.
Je tenais à vous remercier pour cette explication si lumineuse et ces liens très pertinents.🌹
i love you for making such great informative videos thank you
@promptmuse this is awesome. But i have to wait a month for my new pc set up.
Where is the Mac version of this? Can you point me towards the source?
Thanks
Forge is freaking amazing. I just love this tool and it makes things so easier. 😍
well explained!
Cheers dude !
So I got things working and tried using the retro anime lora, and the image comes up black. I copied and pasted from one of the examples, and clicked on the lora to add it, so I'm not sure what I did wrong. Any ideas?
This is interesting! Just make sure you download the LoRa correctly and that it’s got .saftetensor at the end of the file name.
If you’re not logged in and you can right click on the download button, it will save a weird file. That’s not the Lora.
Let me know how it goes , and if indeed this was the issue
Have you tried using Krita AI Diffusion as a graphical interface for SD or Flux? Having experience with graphic design and Photoshop, for me it's the best way to work with generative AI models. I find Forge UI a bit all over the place and not really practical.
Krita is fantastic! To be honest, I would like to do a video or perhaps something a quick comparison between all of the UIs as there are so many now, foocus / krita / comfyui/ invoke/ a1111 etc etc
I think each one serves their purposes well.
I agree, forge interface takes some time to get used to. Transferring over from automatic 1111 it does feel like coming home (without the bugs and smoother performance ).
And then for animation that’s a bit more complex than this video. I would jump into comfyUI ( so I don’t have to pay for a service).
They’re like Pokémon gotta catch them all
oh. I thought it was going to be local animation, but I got Forge installed anyway, which i prefer over comfyui, so thanks :)
any way i can train my own images and use a realistic lora with flux?
Idea for video: using stable diffusion or similar to remove video background
hmmm....everything was going great until sanity check....the model showed int he field...but after hitting generate it disappeared. command window says: Model type not recognized. checked the folder and the model name was now appended with the word "corrupt" at the end...giving it a third try
any update?
You're so 👍 !!! 🎉
❤U!
🤍🤍
Let's go!
Frrrridaaay energy🔥🔥
I wish they would hurry up with the control net support though.
I know I know 🤣🙏
Thanks!!!
You are welcome 🤍
I keep getting this error message! "RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:"
great video thank you
Thank you so much @clydiie 🤍
I Don't know if a "my computor problem" or a "file problem" but unpacking this file takes ALOT of time, it planed to take around 90 hours for me, never ever seen that before 😯
@@geekpotion 90 hours …. No, it should not take that long. Are you zipping? The folder? Is that what’s taking a long time? Should take a few minutes top. Do you have enough space on your hard drive?
Sorry if that came across aggressive I was just doing text to speech whilst leaving the house 🤣
@@promptmuse thanks for responding, it was a me problem, I updated 7zip and the difference was day and night
Will installing forge "break" an automatic1111 install?
Doesn't work for me on a 3090 egpu. cmd says it tries to free up memory but can't.
Great explainer video ❤.
Do I need stable diffusion installed on my system to be able to install Flux?
Also, How to train a Lora on my own images on flux. Could you please also make A tutorial on that.?
Hey, nope you don’t need to stable diffusion at all.
Yeah, definitely will look at making Lora video 🤍
@@promptmuse OK.. Thanks. I see some tutorials where they have mentioned installing the latest version of python for installing flux.
Actually, I have been looking for the easiest way of installing flux, and there are so many tutorials out there with different ways of doing it, that it gets pretty confusing for a newbie like me. 😅
Oh yes, totally understand it can be confusing.
If you install for Forge UI first that actually has the correct python version packaged in with it so you don’t need to do anything else 👍
Hailou is the best! for the moneys worth and time it takes to make them
Recommended settings if I'm running this setup with an RTX 3070? First generation I tried to do after installing the UI and the models was showing an hour plus to create results...
And at the end all I had was a black picture of the size specified. I'm assuming that I really need to get a card with more memory or an even more optimized model?
@@TheGodExperiments It's running fine for me on a 3060 Ti, so I would assume it's something with your settings or installation. Did you change the UI to Flux with the radio button? The sanity check image with her settings took around 2 minutes I'd say. I used the nf4 version.
Possible on AMD GPU ?
It use cuda so not for amd ?
I like your channel and maybe you like what we do.
I myself have been working with AI for almost 3 years, but also with blockchain technology and how it can be used in the film industry.
For video, I really like RunwayML
I've just not had good results, perhaps it 'PINIC'. But for the next video will look into it more, as my subscription accidentally rolled over again for the month (Mybad) ha so I'll play with it a bit more!
@@promptmuse me too. I find Kling a lot more reliable for the time being
@@promptmuseworks well as long as I use one of my images as the first frame. My subscription rolled over too, whoopsie
Pure ref is not free, but pay as much as you want.
Now, if you put in 0 input in the pay as much as you want, it’s free 👍
@@promptmuse I wouldn't encourage not paying for it, especially since it's a good product, that people use a lot. It feels.. dishonest..
@@SolearGnG the devs made it pay what you want so people can use it for free she's bringing in more people to try pureref and anyone who likes it can donate whatever they like - wouldn't it be more dishonest by saying its not free
@@Clydiie No, because it's not. :D It's pay what you want, not free. Saying it is free is factually incorrect.
@@SolearGnG It's not free for businesses, but it is free for personal use. That selected amount is just a donate button. When you enter 0 the total amount will literally say "free." lol
Do you think this will work on 2015 gtx980 8GB ? These only use GPU, rlght? I have an old laptop with i7 2.9 GHZ , GTX980 8 GB GPU and 32 GB RAM. But these do not matter, only GPU, am I right?
Give it ago, I cannot see why not. If you do go for it, just make sure you down the Flux model Nf4(as pointed out in the video). Make sure you got enough airflow for the laptop as it may get a lil warm :)
@@promptmuse thank you for the fast response :) I’ll try and share the results here as well. 🙏😊🤓
@@FunniestCatsandPets good luck, yes I would love to know if you works on your specs.
If you do get any memory issues, you can decrease the weights in the UI. Also makes smaller images and then go about upscaling them after if you run into issues.
Do not run any programs in the background either ( I’m sure you know this all already) 🤣
Looking forward to hearing back from you !
@@promptmuse many thanks for the directions, makes it easier to follow up 🙂 will deliver feedback and thank you again 😊
into pentesting lately? XD
Multiple fingers... more than six. [Me looks at own hand]
You mean you haven’t got 6 fingers per hand ? … how odd !
Quite possibly the most beautiful nerd that ever lived. Makes me want to lift weights and purchase minoxidil 😂 Games Master's 'Games Mistress' with a hint of John Bentley.
Eww.
make your self an Ai girlfriend so all the lonely AI prompter will have something to talk too
it seems when i try to run the sanity check it errors and when i check the checkpoint files i saved in stable-diffusion
flux1-dev-bnb-nf4-v2.safetensors Full flux-dev checkpoint with main model in NF4.
@@blenderisfunty hey, try to re-download the checkpoint file again as sometimes it can become corrupt. It takes probably 45 minutes to download in total, sometimes longer depending on spec. Also, if there is a Wi-Fi interruption that can corrupt the file too 🤞 good luck let me know if this is a fix for you
@@promptmuse Iin firefox i used save link as and it didnt download anything so i clicked into the hugging face and downloaded from there to my stable-diffusion folder, now no files are corrupted but i i get an
error
Connection errored out.
pop up toast on the webpage
looks like this is where the command prompt gets hungup:
[Unload] Trying to free 7725.00 MB for cuda:0 with 0 models keep loaded ... Done.
[Memory Management] Target: JointTextEncoder, Free GPU: 3290.45 MB, Model Require: 5154.62 MB, Previously Loaded: 0.00 MB, Inference Require: 1024.00 MB, Remaining: -2888.17 MB, CPU Swap Loaded (blocked method): 3486.38 MB, GPU Loaded: 1740.62 MB
Moving model(s) has taken 2.85 seconds
Distilled CFG Scale: 3.5
Press any key to continue . . .
looks like this was a firefox issue as i reran it in edge and got a response image!
I’m still troubleshooting some speed issues since it’s currently taking me about three minutes per generation.
However, I found that downloading the Flux file directly from lllyasviel (Hugging Face)-instead of using the "copy link" steps-helped me avoid the "corrupted file" issue entirely.
Failed to recognize model type. On first run :( Saying something about webui/backend/loader line 272 is giving error
Hey, which model did you download? Sometimes if the model didn't completely download correctly (for whatever reason) and the file is corrupted, this can be an error. This could be the issue, if you retry the download again and let me know if the issue persists.
@@promptmuse I think that's exactly what went wrong. I guess the save link option doesn't work quite the same in Opera. I see now that the models downloading are much larger. My oversight. Thank you! I love learning from you!
@@CorinneIsIn Oh good ! They take enough time to have a coffee break and a twitter browse...
@@CorinneIsIn i am using Opera, too and experiencing this same problem. Let me switch to Chrome and see if that works!
@@agente_ai LOL I actually just allowed it to go to the hugging face link and clicked "download" and it's working great now! Opera is amazing but has its quirks for sure! Here's to FLUXing! Woot!
Sorry, great video but didn't work for me. Couldnt find the "sanity checks". Step number three just has a link to PUREREF and CIVIT AI. It wasnt even following my prompt, A prompt for an astronaught was turning out a grainy black and white portrait of a woman. Going back to SDXL. at least that worked.
Hey sorry to hear this, was the UI switch over to flux? If you still want to try, it might be worth trying to download the flux model again. The file can become corrupt if wifi is interrupted or incomplete download. You can use your SDXL in forge as well and switch back and forth.
yup. keeps saying not recognized and model file name changes from "flux1-dev-bnb-nf4-v2.safetensors" to "flux1-dev-bnb-nf4-v2.safetensors.corrupted"
Hey, just make sure that the file is probably downloaded, it does take some time as it quite big.
Also, sometimes the file becomes corrupted if let’s say the Wi-Fi is interrupted and will give this error.
I've tried this several times and it just gave a 39kb file. click on the link to go to the page then where you see "This file is stored with Git LFS . It is too big to display, but you can still download it." click on the download link to start downloading the 12GB file.
Please keep up the great work, brilliant tutorial.
Thank you 🤍🙏