Join the conversation on Discord discord.gg/gggpkVgBf3 You can now support the channel and unlock exclusive perks by becoming a member: pixaroma th-cam.com/channels/mMbwA-s3GZDKVzGZ-kPwaQ.htmljoin Check my other channels: www.youtube.com/@altflux www.youtube.com/@AI2Play
Once again, I came here for the "topic" and I learned 5 other things. POWER LORAs with details from CIVITAI??? Classic! Great job as always! Thank you!
This version of LoRa loader is amazing, i don't have to go back to the website to check the trigger words and can manage multiple LoRa at the same time. Btw, i also want to suggest content for your consideration on how to create consistent character using character reference/sheet that can be done through controlnet that captures different angles and expression of a character. I saw it somewhere but forgotten. It would be really interesting to see your way of doing it.
loving your vids, very clear info! I have to go through them all, you might have covered it already but I was wondering if the q8-gguf is generally better than the nf4-v2?
I dont use nf4 anymore, since with gguf, even the node creator recommends gguf "NOTE: This is very likely Deprecated in favor of GGUF which seems to give better result" github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4 so for me gguf is almost like dev fp16 in quality, but faster
Hi mate great video, just when I was researching lora lol, your video popped up, quick question, if I train 3 characters can I load all 3 in that workflow? I mean I want 3 characters all to be in one image, I'm guessing I just keep adding loras using the plus button to add more loras? Please let me know thanks.
They will get mixed faces because will take something from each lora. Only if you train one lora maybe with a group of same 3 people. Or use inpaint with a single lora to change one face at the time
I have used 2 Lora's together, e.g. myself and either Mona Lisa or Lara Croft. 80-90% of the time it was able to get both faces separately in the same image.
@@SebAnt thats fantastic! im currently started training my first character, not sure how long it will take with an rtx 4070, but i hope to do all 3 by tommorow so i can test it out. thanks again.
@ thanksss , I would like to see a tutorial for my own objects. I think it would be a good idea, because for characters or people there are quite a few.
well i didnt get any error using it, but the effect doesnt seems to be so strong like on dev, so maybe depends on lora, but at least i didnt got an error
I have two LoRA models representing two different people, but I want to include both of them in one image. Is there a way to do that? I look forward to receiving your response, sincerely thank you.
when you mention the trigger word in prompt add details on how it looks, like if you have triggerword1 for one lora and triggerword2 for second do like photo of a triggerword1 man wearing a hat and a suit, next to a blonde triggerword2 woman, wearing a red dress. So more details the better will understand which is which, so you still have some mixing but you get better chance to get both in the image.
@@themrkurin I didn try it but probable can use some nodes that segment the image so you can use one descript for one side and another description for other area, search for words like sam, segs, segment anything
@@radedr007 join button should appear here is direct link th-cam.com/channels/mMbwA-s3GZDKVzGZ-kPwaQ.htmljoin for questions you can join discord discord.gg/gggpkVgBf3
Your ComfyUI / Flux tutorials are so good! Thank you for sharing your knowledge! Do you have, or would you consider making a tutorial on training your own Flux Lora styles on your local machine?
I didn't find an easy way to use it :( there is fluxgym but still have some errors on some systems, the best way for me was still online with a few dollars
![BOWKYLION - วิงวอน (ex-change) [Official MV]](http://i.ytimg.com/vi/cLdnZWDaO-w/mqdefault.jpg)
Join the conversation on Discord discord.gg/gggpkVgBf3
You can now support the channel and unlock exclusive perks by becoming a member:
pixaroma th-cam.com/channels/mMbwA-s3GZDKVzGZ-kPwaQ.htmljoin
Check my other channels:
www.youtube.com/@altflux
www.youtube.com/@AI2Play
The show info tip to get to the trigger words is golden! Thank you so much!
Yeah is very useful ☺️
Thanks, this is one of the BEST episodes yet!
Thank you so much ☺️
no doubts, thats make it clear a lot of things
Once again, I came here for the "topic" and I learned 5 other things. POWER LORAs with details from CIVITAI??? Classic! Great job as always! Thank you!
Glad it helps ☺️ thanks for support
Thanks for this tutorial I Learnt a lot and it reconfirmed a few things I was not aware of about. The power lora is fantastic.
Glad it was helpful 🙂
Literally just today I decided to finally figure this out and here's this video waiting for me, perfect
Thanks, that option to show lora trigger word(s) from a LoRA file is neat feature.
A detailed and accurate description of a complex method to understand! Thank you
Cheers to legends and special cheers to king seb 👑
Glad it helped ☺️
This version of LoRa loader is amazing, i don't have to go back to the website to check the trigger words and can manage multiple LoRa at the same time. Btw, i also want to suggest content for your consideration on how to create consistent character using character reference/sheet that can be done through controlnet that captures different angles and expression of a character. I saw it somewhere but forgotten. It would be really interesting to see your way of doing it.
great video, as always and really showed how to get the best out of flux with Loras. Thank you very much 🙂
Amazing content. Loved your series. Please keep it up. Could you create an episode on Comfy UI Flux LoRA training?
Once i find an easy method, locally was a little bit difficult to install, online was easy but cost a few dollars
Thanks!
thank you so much Marcel 🙂
Thanks a lot, very well explained
loving your vids, very clear info! I have to go through them all, you might have covered it already but I was wondering if the q8-gguf is generally better than the nf4-v2?
I dont use nf4 anymore, since with gguf, even the node creator recommends gguf "NOTE: This is very likely Deprecated in favor of GGUF which seems to give better result" github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4 so for me gguf is almost like dev fp16 in quality, but faster
Make more videos on different - different LORAS
Hi mate great video, just when I was researching lora lol, your video popped up, quick question, if I train 3 characters can I load all 3 in that workflow? I mean I want 3 characters all to be in one image, I'm guessing I just keep adding loras using the plus button to add more loras? Please let me know thanks.
Basically I want 3 consistent characters in a short 3d animation
They will get mixed faces because will take something from each lora. Only if you train one lora maybe with a group of same 3 people. Or use inpaint with a single lora to change one face at the time
Ok thanks ate seems this is one area that is hard to crack. I would love you to try one day and a video would be great 👍, thanks
I have used 2 Lora's together, e.g. myself and either Mona Lisa or Lara Croft. 80-90% of the time it was able to get both faces separately in the same image.
@@SebAnt thats fantastic! im currently started training my first character, not sure how long it will take with an rtx 4070, but i hope to do all 3 by tommorow so i can test it out. thanks again.
😎Danke!
Thank you for support ☺️
Thanks Bro
Sorry, could you recommend some comfyui workflow to use a lora that was trained to recognize an object of my own brand?
Should work with any workflow with Lora for flux, i have some in the pixaroma-workflows on discord
Check discord channel the pixaroma-workflows on the pixaroma.server
@ thanksss , I would like to see a tutorial for my own objects. I think it would be a good idea, because for characters or people there are quite a few.
How do you have all your PC live metrics on display at the top of the screen?
There is a custom node you can install from manager, custom nodes manager, the node :crystools
imho - Lora Loader from Custom Script is more convenient to use. Give it a try)
hey, so i have the flux dev nf4 by bitsandbytes, it doesn't seem to work with lora, is there a workaround or fix ? much appreciated!
Try maybe gguf models, those works ok, i only used nf4 for a few days before gguf appeared
Can be use same LoRa model with both Flux Dev and Schnell?
well i didnt get any error using it, but the effect doesnt seems to be so strong like on dev, so maybe depends on lora, but at least i didnt got an error
I have two LoRA models representing two different people, but I want to include both of them in one image. Is there a way to do that?
I look forward to receiving your response, sincerely thank you.
when you mention the trigger word in prompt add details on how it looks, like if you have triggerword1 for one lora and triggerword2 for second do like photo of a triggerword1 man wearing a hat and a suit, next to a blonde triggerword2 woman, wearing a red dress. So more details the better will understand which is which, so you still have some mixing but you get better chance to get both in the image.
@ Is there a way to load LoRA 1 into a designated position 1, LoRA 2 into designated position 2, and run them together to create a complete image?
@@themrkurin I didn try it but probable can use some nodes that segment the image so you can use one descript for one side and another description for other area, search for words like sam, segs, segment anything
@@pixaroma thank you for your help
at 16:06 how did you open cmd comfy which shows the time consumed?
When you run comfyui automatically is open a command window on the task bar that is minimized so you just click on it on taskbar to open
@@pixaroma im using comfy in pinokio
@@pixaroma I'm using comfy in pinokio and it doesn't show anything
and how can i become a member of this channel , I cant see any join button.
@@radedr007 join button should appear here is direct link th-cam.com/channels/mMbwA-s3GZDKVzGZ-kPwaQ.htmljoin for questions you can join discord discord.gg/gggpkVgBf3
The best teacher! Thank you for your lessons and knowledge!
You are welcome ☺️
Your ComfyUI / Flux tutorials are so good! Thank you for sharing your knowledge! Do you have, or would you consider making a tutorial on training your own Flux Lora styles on your local machine?
I didn't find an easy way to use it :( there is fluxgym but still have some errors on some systems, the best way for me was still online with a few dollars
@@pixaromaThanks!
hum why my UI new look difference >.< queue is stick on top
Not sure maybe they updated today with a new one, or you have a different version or settings
Ahh, went he toggled to Light Mode it was like staring at the sun, why is that even an option! lol
😁
Great video. I install few add ons in FORGE UI and i cant see them. Any idea why? Thanks
not sure, I dont use forge since I learned comfui, maybe is because of the new version they keep updating and fixing bugs
Thanks!
Thank you ☺️