Hey Olivio! Thanks so much for the awesome review - really means a lot. I’m thrilled that you and your audience are enjoying the LoRA, and it’s great to see the community finding value in it. Seeing feedback like this genuinely motivates me to keep refining and improving. Your insights and guidance will definitely help me as I work on future versions. Appreciate you taking the time to showcase it - it’s a big encouragement for me to push things even further. Cheers! 😉
Not really =)) It might look Eastern European-heavy just from the Civit samples, but that’s just one slice of the aesthetic. My dataset’s actually a mix, with plenty of U.S. scenes thrown in, like trailer parks and gritty American vibes too
Maybe I'm too old-fashioned or too used to old cameras and photos from the early 2000s, but to me this photo 2:35 looks 100x more realistic than this next one 2:45
Is there a way to inject a Pose into this? a face as the base image and then the pose from there. I find with all the Pulid tutorials the skin is always plastic and unrealistic not sure if this would be able to be adapted.
Yeah, I experimented a lot with pulid + my lora. Just play with settings, for example start point 0.2 and end is 0.9, also try different fusion options
i have an issue with forge UI, the UI is slow, when moving cursor above options theres lag there. generations are fast but the UI is laggy, does it occur to you ? how to fix it ? can you make a short video about it
Is there any way to get grain into the images to make them more 'film like'? They are always too perfect, and how can I avoid to create a flash image to enforce grain or underexposure? Putting 'flash' into the negative prompt obviously has no effect
If you want them a lot more photo like, i would suggest to edit them in photoshop or lightroom to give them some actual photo filters. pluss they have a noise function in there too. If you use photoshop, try the "camera raw" filter
What is best local upscaler model for upscaling thousands of Flux realistic landscape images? Topaz and others cost a bank at many places, they are very expensive.
I don't see the difference between the Ultra Realistic Flora Project and Amateur Photography [Flux Dev] with a power of 0.8 settings. What was the advertisement for?🤔🤨
Still waiting for a replacement to 1.5 or SDXL, Flux is too big and too slow. We need something that's smaller and faster, not bigger and bulkier. The 4GB SDXL model is a step in the right direction, why doesn't that model get the attention?
Well the SDXL model still has some attention, but as time moves on and people want better outcomes, I guess its just logical that the model sizes grow in proportion. More details and complex images need more data.
@MrMsschwing Not quite.. 1.5 still has some attention. SDXL is still the main preference for most people if you simply look at the number of downloads for each model. If you thought Flux is popular because of all the uploads, you've been fooled because what the model authors does is just throwing out different quantisations of the same model, which makes Flux appear popular. Also, by releasing the same Flux model but in various quantisations the authors can abuse Civit's Buzz system, and Civit is fully aware of the situation and are working on a solution. So no, Flux is not popular among anyone but those who have 16GB cards or above, and the model authors themselves.
@MrMsschwing what I'm trying to say is... Everyone wants a better model than SDXL but they also want a model they can run. Flux is too taxing. If you know where the sweetspot is, its 12GB vram and below.
@@SteamrickI want to know what model he's using because I've been having trouble getting any LoRA to work correctly for Flux. There are so many different versions and it's gotten confusing.
@@WesTheWizardOne of the reasons I dont use Flux is the word-soup, and because of how big and slow it is. I recommend checking out the 4GB SDXL model, and perhaps help expand that model.
Not bad until plugin PulID after that the sign text is diminished. Just started testing it out but I'm guessing it's a compromise in dominance between pulID and lora, or flux, lora and pulid? Then it's a nice clone vs nice text vs realism. In hindsight the sign is relatively easy to fix seperately.
#### Links from my Video ####
GET MY WORKFLOW: www.patreon.com/posts/ultrarealistic-115566041
UltraRealistic Lora: civitai.com/images/38271885
👋 hi
Hey Olivio! Thanks so much for the awesome review - really means a lot. I’m thrilled that you and your audience are enjoying the LoRA, and it’s great to see the community finding value in it. Seeing feedback like this genuinely motivates me to keep refining and improving. Your insights and guidance will definitely help me as I work on future versions. Appreciate you taking the time to showcase it - it’s a big encouragement for me to push things even further. Cheers! 😉
congrats!
everything looks like its from Eastern Europe with this Lora
Not really =)) It might look Eastern European-heavy just from the Civit samples, but that’s just one slice of the aesthetic. My dataset’s actually a mix, with plenty of U.S. scenes thrown in, like trailer parks and gritty American vibes too
Really enjoy your videos! I feel the moment I get stuck on something you have a video on it. Perfect timing every time!
I'm getting OK results, but I still think the Lora "Amateur Photography [Flux Dev]" is the best for this type of stuff.
Thanks for suggestion and you right its better ,with my settings and Lora weight 0.85 its just perfect.
Its time to get Laid? What? Alright then!
this is not realistic lol
@@koufdell You're saying its not realistic that I could get laid?
it´s so realistic!! Thanks!!!
Maybe I'm too old-fashioned or too used to old cameras and photos from the early 2000s, but to me this photo 2:35 looks 100x more realistic than this next one 2:45
How did you get the backroomcastingcouch scenery in that last image?😅🤣
You're doing 0.87 as the starting point for all LoRAs or for this particular LoRA?
Hahahaha. Kicked it off with a…wait for it…bang.
your workflow in NICE, but the particular lora - is just OK
Is there a way to inject a Pose into this? a face as the base image and then the pose from there. I find with all the Pulid tutorials the skin is always plastic and unrealistic not sure if this would be able to be adapted.
Yeah, I experimented a lot with pulid + my lora. Just play with settings, for example start point 0.2 and end is 0.9, also try different fusion options
You said Simple Advanced Ksampler.. hehe well it is late time, hehe
Thanks for video
Please help me to understand whether Euler is worse than DPM in Flux?
i have an issue with forge UI, the UI is slow, when moving cursor above options theres lag there. generations are fast but the UI is laggy, does it occur to you ? how to fix it ? can you make a short video about it
please some videos about Arch viz in flux!
Is there any way to get grain into the images to make them more 'film like'? They are always too perfect, and how can I avoid to create a flash image to enforce grain or underexposure? Putting 'flash' into the negative prompt obviously has no effect
If you want them a lot more photo like, i would suggest to edit them in photoshop or lightroom to give them some actual photo filters. pluss they have a noise function in there too. If you use photoshop, try the "camera raw" filter
Is there an other way to use this Lora V2 than than with this "spaghetti box" ? Thanks.
What is best local upscaler model for upscaling thousands of Flux realistic landscape images? Topaz and others cost a bank at many places, they are very expensive.
I don't see the difference between the Ultra Realistic Flora Project and Amateur Photography [Flux Dev] with a power of 0.8 settings. What was the advertisement for?🤔🤨
kind of like how nike and adidas both make sneakers and both have their fans...
Where do I find that "FaceDetailer"?
when you look on top of every node, it says the pack name. in this case the Impact Pack for Comfyui :)
There are sooooooooooo many options out there it has become confusing!
Still waiting for a replacement to 1.5 or SDXL, Flux is too big and too slow. We need something that's smaller and faster, not bigger and bulkier.
The 4GB SDXL model is a step in the right direction, why doesn't that model get the attention?
Well the SDXL model still has some attention, but as time moves on and people want better outcomes, I guess its just logical that the model sizes grow in proportion.
More details and complex images need more data.
@MrMsschwing Not quite.. 1.5 still has some attention. SDXL is still the main preference for most people if you simply look at the number of downloads for each model.
If you thought Flux is popular because of all the uploads, you've been fooled because what the model authors does is just throwing out different quantisations of the same model, which makes Flux appear popular.
Also, by releasing the same Flux model but in various quantisations the authors can abuse Civit's Buzz system, and Civit is fully aware of the situation and are working on a solution.
So no, Flux is not popular among anyone but those who have 16GB cards or above, and the model authors themselves.
@MrMsschwing what I'm trying to say is... Everyone wants a better model than SDXL but they also want a model they can run. Flux is too taxing. If you know where the sweetspot is, its 12GB vram and below.
is this a ultra realistic lora or a goth/depressing scenes lora?
But its for Non Commercial? That is bad
What model (and type ? unet/gguf?? ) is used as the base model ? And how much VRAM is needed ?
This is what I want to know!
It's a lora. Unless the lora description page specifies, feel free to use anything that's technically compatible and see what the result looks like.
@@SteamrickI want to know what model he's using because I've been having trouble getting any LoRA to work correctly for Flux. There are so many different versions and it's gotten confusing.
@@WesTheWizardOne of the reasons I dont use Flux is the word-soup, and because of how big and slow it is.
I recommend checking out the 4GB SDXL model, and perhaps help expand that model.
Not bad until plugin PulID after that the sign text is diminished. Just started testing it out but I'm guessing it's a compromise in dominance between pulID and lora, or flux, lora and pulid? Then it's a nice clone vs nice text vs realism. In hindsight the sign is relatively easy to fix seperately.
this lora changes the faces of the characters. Avoid
first
Really? Havent seen this kind of comment since around 2010 xD