Installing this manually without PiNO will give you much faster and more powerful performance. It's easy to install, about the same as Pino; there is a 2-click install process out there, and it's on YT somewhere, but the manual is easy, too. It's faster, requires less RAM, and will function without an NVID card if you use the right 2-click installer. Also, some of the features you don't notice changes, such as you will be outside of the Pino installation. BTW I love pino, but sometimes its in the way of the resources and the func pattern and paths of choice. Thanks for your videos, I enjoy them, peace.
Just installed it without pino, it has to download 6.4 gbs of data every time you change model/style.. Folder was 36 gbs before I got anything I liked.
This seems way better than my technique, thank you! I used a really long description of the woman for the positive prompt, so it couldn't possibly render anyone else. It mostly works. I made "Thelma Clickbait" to add to my thumbnail images, it actually works too 😁
I have not made money with any Ai influencers if that's what you meant. But I do have this yt channel monetized, which generates a small amount of money.
Introduction to Focus Tool - 00:00:01 Installation and Basic UI Overview - 00:00:31 Creating a Base Image - 00:02:02 Editing and Refining Images - 00:03:33 Advanced Settings and Features - 00:05:00 Conclusion and Additional Resources - 00:07:23
I have it open and it says terminal, when i click default it starts downloading something out of 6.62GB is this normal or am i meant to enter through a different path?
this was awesome, can you go more in deep on the lora and other options?, im trying to do as you say, but i dont know if i have to use juggernaut or another model!, thanks in advance
Bro what TF kind of version you told us to download man? It took me over 30gb and it doesn't even have Juggernaut Xl. It has realisticStockPhoto v20 as base model and refiner... The pictures look bad
Thanks for your video, really useful! I have a question: is there a way to make an AI generate subject holding another object taken from a picture? For example, I've generated a character, now I want him to hold a particular videogame controller. Is there a way to make this happen using a controller photo as input? Thanks
Thanks! It's funny you mention game controllers as I just tried that recently and most of the controllers in the images I got, were facing the camera while the person was "playing". I have not seen any controlnet model that would let you replace objects or let you put certain objects into images in Stable Diffusion. So to my knowledge this would still require traditional image editing sw on your pc or mobile.
Thanks, if you are having trouble getting the face to look right, you could try increasing the "weight". And make sure to use a similar prompt like the one that was used to create the base image. It seems to me though that it works better with some faces compared to others.
When I am handing over the fees and handing over someone else's body to him, then only the photo etc. is being generated, it is of clothes or the face etc., the whole face is not being visible. Can you tell me the solution for this?
thank you so much for the amazing tutorial! Is there a way to generate different camera angles of the base image? I've tried to use image prompts with various face angles but this messed up the looks and produced inconsistent face (didn't look like the same person)
Thanks! What you can try is using a second image for the pose or camera angle that you want (similar to the bodybuilder in the video). It might take a few tries to get the face right and it might not work on all angles. Hope that helps
Hi, thanks! You can try adding the same word (e.g. "sad") to the prompt twice or three times even.Once at the start of the prompt, in the middle, and in the end. If that doesn't help, try lowering the "Stop at" value a bit. Hope that helps
Thanks for the video! When i try image promt feature and I selected an image, it gives me an error. It doesn't download anything. I'm getting this error: "File "C:\Users\Particular\pinokio\api\fooocus.pinokio.git\Fooocus\env\lib\site-packages\torch\cuda\__ init__.py", line 289, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled "
I didn;t had an option otherthan subscribe more details video even though i have been using it for a month now, its more informative thanks m8, seeems ai generated voice :D
If you only want to change the hair of someone (or if you want to keep only the face and change everything else in the image), you can try to use inpainting and highlight what you want to change.
FaceSwap does not generate 100% when I tried with real person photo (closed up). It is like 70% accurate. I have tried many times to make output as close as possible to the original photo, but no luck. And Fooocus makes the output little younger than actual pictures. If anyone knows better settings with real photo, Please let me know!
Yeah I think it doesn't work that well with photos is because they were not created with the SDXL checkpoint that is being used to generate the new image. But I'd also love to hear some tricks in case anyone knows any
I see your one takes about 20 seconds to generate the image but mine one takes around 3 minutes with the default settings. What do you think influences it the most? The graphic card?
Yes it mainly depends on your GPU. If you're interested, I made a video about PC hardware to run local Ai tools: ruclips.net/video/yhIvOxqtvJc/видео.html
On an iMac from 2019, from what I can tell, you are fully running this on an a CPU which makes this so slow. It runs faster on a Mac with M1/M2 chips
6 месяцев назад
@@AiVOICETUTOR I'm getting around 25 minutes per image using the quality performance setting, on an M1 Pro with 16GB of RAM. Any tips to improve performance? I'm running it on CPU and GPU at the same time, so it's maxed out I guess.
I think this would require another SDXL controlnet which doesn't exist yet to my knowledge. Maybe you can add your brand logo via traditional image editing sw on your PC or mobile.
You could try running Fooocus online via Google Colab but I have no idea how well this runs in a mobile browser. More info here: github.com/camenduru/Fooocus-colab
When i try image promt feature and I selected an image, it gives me an error. It doesn't download anything. Downloaded from Pinokio and I tried in Pinokio browser as well as external Firefox browser as well.
Thanks for the reply. Now its working. I was trying to upscale an image. And by mistake I was selecting already upscaled image which was bigger in size.@@AiVOICETUTOR
@@AiVOICETUTOR i was vastly disappointed with the results. Unless if it gets better with extreme tweaking and crafting. But i had better results with free online algorithms like perchance than this setup.
Hey sir, I'm getting errors while using the tool. If i faceswap, then go to upscale or unpainted the pic, it shows error. Meaning, i couldn't able to multiple task back to back. To do the next work, i have to again close the page and then again run the process. Can you tell me how to encounter it?
@@AiVOICETUTOR i was talking about fooocus google colab version. But i installed it via pinokio too, but it directly shows an error, it doesn't specify what error it is. But not a single image was generated in the local server.
Hey you can try adding something like "highly detailed background" in the prompt and to me it seems to be a bit easier to get non blurry backgrounds when the aspect ratio is sth like 1/1 or 16/9 and not a portrait but its certainly hit n miss. Hope that still helps
Yeah it should work but you might need to do some extra steps if it doesn't work out of the box via Pinokio: github.com/lllyasviel/Fooocus#windowsamd-gpus
If you cannot get it via the prompt alone, you can use a second image as input for the pose (like I did with the bodybuilder in the video). If that person is shown from head to toe and you make sure to use the same or a similar aspect ratio as the image with the pose, it will work.
I installed the program. I uploaded Lora and the model, but it shows 35 minutes to produce even the easiest prompt like "girl". What can i do@@AiVOICETUTOR
There's a few reasons why it's taking longer than the usual stable diffusion: It's an SDXL checkpoint and when using the image prompt, fooocus is basically running an SDXL controlnet and it's all done in quite a high resolution with a LoRA enabled.
@@AiVOICETUTOR And how is it on the other hand in Stable Diffusion? (And which webUI? Do you know others? I think Comfy could be more advanced than Fooocus.)
Do you realize it's an AI voice explaining how to create an AI influencer? It's like a vicious circle. Soon, AI will be everywhere, but well, we can't do anything about it; it will simply tell us what to watch and how to make it. Modern life, isn't it?
Sorry its not working for you. I hope it's just a temporary error with the Fooocus installation and that it will be fixed shortly. You can try updating Pinokio (by installing the latest version from the website)
Hey! I'm wondering if you know how to make the AI influencer version, but in a detailed background. I can never get that, every one I generate is with a blurry background unfortunately
Hey, I hadn't noticed that yet but thats a good observation! My first hunch would be to deactivate the style presets and see if that helps. I'll have a look myself the next time I'm on my PC
@@AiVOICETUTOR I've tried alot so far! Trying other lora's like better details, and realistic vision. Tried other models/settings/weight! The only time I can sort of get a less blurry, but still not satisfying result. Is when I put the person in inpaint, then select the background, and put the prompt as in highly detailed background. That sort of makes is 5/10% less blurry. I know it must be possible, because it can do both perfectly! Just not together haha
I've played around with a bit now and while I did get some pictures with like a 95% clear background, I have no idea how to reproduce it and it seems a bit random. I'm not sure if i'm imagining it, but it seems a bit easier to get non blurry backgrounds when the aspect ratio is sth like 1/1 or 16/9 and not portrait
@AiVOICETUTOR yes! It's very strange, it's mostly focused on a portrait. Even when you put portrait, blur, blurry etc as negative prompts! I will try out some other aspect ratio's later today thanx
Seriously, this is the best, shotest, most concise explanation so far on this topic. Thanks for this.
Thank you so much for taking the time to post your comment. You just made my day!
Installing this manually without PiNO will give you much faster and more powerful performance. It's easy to install, about the same as Pino; there is a 2-click install process out there, and it's on YT somewhere, but the manual is easy, too. It's faster, requires less RAM, and will function without an NVID card if you use the right 2-click installer. Also, some of the features you don't notice changes, such as you will be outside of the Pino installation. BTW I love pino, but sometimes its in the way of the resources and the func pattern and paths of choice. Thanks for your videos, I enjoy them, peace.
Where? Can you send me another tutorial? My PC is running so slow when I use it 😂
Just installed it without pino, it has to download 6.4 gbs of data every time you change model/style.. Folder was 36 gbs before I got anything I liked.
This seems way better than my technique, thank you! I used a really long description of the woman for the positive prompt, so it couldn't possibly render anyone else. It mostly works. I made "Thelma Clickbait" to add to my thumbnail images, it actually works too 😁
Cool! I'm glad the video was helpful to you!
That´s a very good explanation, short but all the important informations were included.
Have you been able to make money with AI?😊
I have not made money with any Ai influencers if that's what you meant. But I do have this yt channel monetized, which generates a small amount of money.
Introduction to Focus Tool - 00:00:01
Installation and Basic UI Overview - 00:00:31
Creating a Base Image - 00:02:02
Editing and Refining Images - 00:03:33
Advanced Settings and Features - 00:05:00
Conclusion and Additional Resources - 00:07:23
Aiva Rivera is the most beautiful AI model in my opinion ❤
where do you get it?
Thank you, thank you so much, great detais, great infos, thank you again
Thanks for your comment. I'm glad you like it!
I have it open and it says terminal, when i click default it starts downloading something out of 6.62GB is this normal or am i meant to enter through a different path?
this was awesome, can you go more in deep on the lora and other options?, im trying to do as you say, but i dont know if i have to use juggernaut or another model!, thanks in advance
Great video. Thank you for the tutorial. Awesome
You're welcome and thanks for your comment!
Hey. Quick question... how do you uninstall fooocus or update to better models when released??
Best video I've seen in this mount
What things can use to help me come up with prompts?
very good
Thanks
Great Work! Thank you
Thanks! I'm glad you like it
Great Video
Glad you enjoyed it
My foocus is not starting it gets to press any key to continue and when I do so it just closes any advice on how to fix?
hey man, can you please make a tutorial on how to load Safetensors Model in Fooocus? that would be really helpful..great tutorial btw!
Just put them in checkpoints folder.
Yeah just put the models in the /checkpoint folder in your fooocus folder.
Bro what TF kind of version you told us to download man? It took me over 30gb and it doesn't even have Juggernaut Xl. It has realisticStockPhoto v20 as base model and refiner... The pictures look bad
Hi do you know how to create dark skinned models. It seems its a bit challenging for AI
Thanks for your video, really useful! I have a question: is there a way to make an AI generate subject holding another object taken from a picture? For example, I've generated a character, now I want him to hold a particular videogame controller. Is there a way to make this happen using a controller photo as input?
Thanks
Thanks! It's funny you mention game controllers as I just tried that recently and most of the controllers in the images I got, were facing the camera while the person was "playing". I have not seen any controlnet model that would let you replace objects or let you put certain objects into images in Stable Diffusion. So to my knowledge this would still require traditional image editing sw on your pc or mobile.
@@AiVOICETUTOR I've got your same issue! Thx for the reply, I will go with traditional editing
great video, however my biggest hurdle so far is replicating the exact same face in different poses and setting, and tips on how to achieve this?
Thanks, if you are having trouble getting the face to look right, you could try increasing the "weight". And make sure to use a similar prompt like the one that was used to create the base image. It seems to me though that it works better with some faces compared to others.
ill give this a go, if you could make a video on custom LORA training that would be dope to see as well@@AiVOICETUTOR
This tutorial provides valuable insights, but I'm curious about its financial significance. What's the monetary value behind it? Uses??
There are people who earn money with the Ai influencers they built. And they make money the same way that real influencers do.
When I am handing over the fees and handing over someone else's body to him, then only the photo etc. is being generated, it is of clothes or the face etc., the whole face is not being visible. Can you tell me the solution for this?
Try using a different aspect ratio.
thank you so much for the amazing tutorial!
Is there a way to generate different camera angles of the base image? I've tried to use image prompts with various face angles but this messed up the looks and produced inconsistent face (didn't look like the same person)
Thanks! What you can try is using a second image for the pose or camera angle that you want (similar to the bodybuilder in the video). It might take a few tries to get the face right and it might not work on all angles. Hope that helps
thank you for the tutorial, can you please recommned the best budget GPU for this kind of projects?
Sure, check out this video or the video description for my minimum PC recommendations for Ai stuff: ruclips.net/video/yhIvOxqtvJc/видео.html
i have no gpu and i can still use this..I use google colab to run foocus
Very cool tutorial and voice.
Thanks! Appreciate the comment
Hello! Can we keep the same face with different expressions (sad, happy, worried)? Great video!
Hi, thanks! You can try adding the same word (e.g. "sad") to the prompt twice or three times even.Once at the start of the prompt, in the middle, and in the end. If that doesn't help, try lowering the "Stop at" value a bit. Hope that helps
@@AiVOICETUTOR thanks a lot for your instructions! I'll certainly try all the suggestions you gave me 😘
Tangs Mr. AI
You're welcome 🫡
Thanks for the video! When i try image promt feature and I selected an image, it gives me an error. It doesn't download anything.
I'm getting this error:
"File "C:\Users\Particular\pinokio\api\fooocus.pinokio.git\Fooocus\env\lib\site-packages\torch\cuda\__
init__.py", line 289, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled "
Hi, are you running this on a windows PC with NVIDIA GPU?
@AiVOICETUTOR Great video, but im also getting this same error code, any tips?
How do you get a full body image?
Thank you it was a useful tutorial
Glad it was helpful!
Do you know how i can sharpen the background? :)
I didn;t had an option otherthan subscribe more details video even though i have been using it for a month now, its more informative thanks m8, seeems ai generated voice :D
Thanks I appreciate it! I had to shift my focus on other things at the moment but I will try to make new videos soon.
Bro I tried generating but it gives nothing but a black image everytime any solution?
I like how Undetectable AI Human Auto Typer works, its a chrome extension that will turn your copy paste context into manually written
Can you make a faceswap that includes hair with this ?
Thanks.
If you only want to change the hair of someone (or if you want to keep only the face and change everything else in the image), you can try to use inpainting and highlight what you want to change.
can you tell me how i can make the hair consistent?
Does this one limit on gpu usage ? can you tell me pls
Everytime I generate it process a little and then shows error , connection errored out . Plz help
Sorry it's not working for you. What exactly does the server window in Pinokio say?
can you use for android this
I want to make a natural picture pls help me more tips
Buenas tardes, tengo un RX 6700XT + Ryzen 5 5600 , como hago para que me funcione Foooocus. O la unica manera es con NVIDIA? gracias
AFAIK to run this on an AMD GPU, you need to use linux. More info here: github.com/lllyasviel/Fooocus?tab=readme-ov-file#linux-amd-gpus
FaceSwap does not generate 100% when I tried with real person photo (closed up). It is like 70% accurate. I have tried many times to make output as close as possible to the original photo, but no luck. And Fooocus makes the output little younger than actual pictures. If anyone knows better settings with real photo, Please let me know!
Yeah I think it doesn't work that well with photos is because they were not created with the SDXL checkpoint that is being used to generate the new image. But I'd also love to hear some tricks in case anyone knows any
i know how to fix it into real face
@@sapphiraph5121how bro
Just tell us @@sapphiraph5121
How? @@sapphiraph5121
i want to generate ***** images how can i do that🙂
Can i still run it if i have lower gpu
Hello. Is there any cloud solution to use the same ai tools in case I have less powerful GPU notebook?
Hi, yes you can run fooocus online in Google Colab. More info here: github.com/camenduru/Fooocus-colab
so, there isn't a "seed" value as in e.g. Dalle-E to reproduce?
It's there but a bit hidden. You need to untick the "random" option: ruclips.net/video/75buuLeT-Q8/видео.html
I see your one takes about 20 seconds to generate the image but mine one takes around 3 minutes with the default settings. What do you think influences it the most? The graphic card?
Yes it mainly depends on your GPU. If you're interested, I made a video about PC hardware to run local Ai tools: ruclips.net/video/yhIvOxqtvJc/видео.html
great video, i have imac from 2019 , i9 32gb ram and 8gb rx580 and takes about 25 mins for a image, that normal? as on paid colab takes 3 mins.
Not normal, try newer or older GPU drivers.
On an iMac from 2019, from what I can tell, you are fully running this on an a CPU which makes this so slow. It runs faster on a Mac with M1/M2 chips
@@AiVOICETUTOR I'm getting around 25 minutes per image using the quality performance setting, on an M1 Pro with 16GB of RAM. Any tips to improve performance? I'm running it on CPU and GPU at the same time, so it's maxed out I guess.
How can someone get in contact with you?
why it use my cpu even i have good gpu please tell me hoe to change
Is it possible to make her wear a shirt with my brand?
I think this would require another SDXL controlnet which doesn't exist yet to my knowledge. Maybe you can add your brand logo via traditional image editing sw on your PC or mobile.
Any tips for Fooocus on how to create face similar to your own but slightly different?
You can use the same prompt that I used and make slight changes to it
How can I add a specific dress I have a photo of to my model's image?
I'm afraid that its not possible yet to add specific items to the images.
I was doing the same, but when I'm doing generate then they show me an error. Can you tell me what problem this is?
What does the error say in the terminal/server window of pinokio?
hello, guess I'm experiencing some technincal issues and I don't know what to do. May I contact you somewhere please? :)
Hi you can post the error here so that I or others can help you
I don't have a computer, the only device I have is a cell phone. Can you use your phone to create an AI influencer?
You could try running Fooocus online via Google Colab but I have no idea how well this runs in a mobile browser. More info here: github.com/camenduru/Fooocus-colab
When i try image promt feature and I selected an image, it gives me an error. It doesn't download anything. Downloaded from Pinokio and I tried in Pinokio browser as well as external Firefox browser as well.
Sorry it's not working for you. What does the error say in the server window?
Thanks for the reply. Now its working. I was trying to upscale an image. And by mistake I was selecting already upscaled image which was bigger in size.@@AiVOICETUTOR
What about the Lama model? It it good for local image production?
Lama is for text generation only
@@AiVOICETUTOR so which version of stable diffusion better
The one that Foooocus uses (SDXL) currently is the best
@@AiVOICETUTOR i was vastly disappointed with the results. Unless if it gets better with extreme tweaking and crafting. But i had better results with free online algorithms like perchance than this setup.
If I get a Macbook M3 pro , can I easily install Comfy UI and work?
There's some info on the official git about running Fooocus on a M1/M2 Mac but I haven't tested it myself: github.com/lllyasviel/Fooocus#mac
@@AiVOICETUTOR Thank you very much 🙏🏻
Hey sir, I'm getting errors while using the tool. If i faceswap, then go to upscale or unpainted the pic, it shows error. Meaning, i couldn't able to multiple task back to back. To do the next work, i have to again close the page and then again run the process. Can you tell me how to encounter it?
Hey what error does it give you in the pinokio server/terminal window?
@@AiVOICETUTOR i was talking about fooocus google colab version. But i installed it via pinokio too, but it directly shows an error, it doesn't specify what error it is. But not a single image was generated in the local server.
great video, but fooocus is better installed without pinokio... :D
Only if you're very familiar with CLI though :p
i use fooocus with 6gb VRAM but without pinokio,, great UI btw.... @@AiVOICETUTOR
My computer rebootet and i cant find the program. How do I open it?
You can launch Fooocus again after a reboot by opening Pinokio and launching Fooocus from the main tab in Pinokio.
hi, when I do a prompt I get error. Does someone knows why I get this error?
Try updating video card drivers
can we add new aspect ratio
Not as far as I know as the tool includes all resolutions that SDXL was trained on
Any advice for get the background more detailed?I really don t like the blur effect on background
Bump
Hey you can try adding something like "highly detailed background" in the prompt and to me it seems to be a bit easier to get non blurry backgrounds when the aspect ratio is sth like 1/1 or 16/9 and not a portrait but its certainly hit n miss. Hope that still helps
@@AiVOICETUTOR I tryed to add "highly detailed background" in the prompt,but nothing changed.
The pinokio isn't downloading
Make sure the download isn't blocked from your web browser.
Is it work with a amd gpu ?
Yeah it should work but you might need to do some extra steps if it doesn't work out of the box via Pinokio: github.com/lllyasviel/Fooocus#windowsamd-gpus
for video?
So far I have not found a video method that looks good enough. Will make a video about it once I do
Does it work on mac ?
Update no it doesn't
There's some info on the official git about running it on a M1/M2 Mac but I haven't tested it myself: github.com/lllyasviel/Fooocus#mac
@@AiVOICETUTOR yes I tried and it doesn't work
The channel "School Of Ambition" is copying your content it seems mate.
Haha interesting! I guess his other videos are stolen from other channels. Thanks for letting me know man!
How can you generate an image of a person from head to toe?
If you cannot get it via the prompt alone, you can use a second image as input for the pose (like I did with the bodybuilder in the video). If that person is shown from head to toe and you make sure to use the same or a similar aspect ratio as the image with the pose, it will work.
@@AiVOICETUTOR thank you, i will try it
Is there a way to keep a consistent bra size?
You could try specifying the size (or measurements) in the prompt.
How much gb data this requires
33GB
how can i add model and lora
Download the model and lora manually and then move them to the models and Loras folder in the Foooocus folder
I installed the program. I uploaded Lora and the model, but it shows 35 minutes to produce even the easiest prompt like "girl". What can i do@@AiVOICETUTOR
You can get a faster GPU. I made a video about PC hardware here: ruclips.net/video/yhIvOxqtvJc/видео.html
Does it free or paid 9
100% free
Why does it take so much time to generate one image?
There's a few reasons why it's taking longer than the usual stable diffusion: It's an SDXL checkpoint and when using the image prompt, fooocus is basically running an SDXL controlnet and it's all done in quite a high resolution with a LoRA enabled.
@@AiVOICETUTOR And how is it on the other hand in Stable Diffusion? (And which webUI? Do you know others? I think Comfy could be more advanced than Fooocus.)
Audiogradio
Do you realize it's an AI voice explaining how to create an AI influencer? It's like a vicious circle. Soon, AI will be everywhere, but well, we can't do anything about it; it will simply tell us what to watch and how to make it. Modern life, isn't it?
Hah good point. I guess soon you'll simply tell the Ai "make a tutorial explaining how to create an ai influencer"
An error immediately after installation on the first run, great
Sorry its not working for you. What does the error say?
@@AiVOICETUTOR it is an AMD Problem. the solution is some changes in the run.bat
@@wandtellermanncould you please tell me what should I change there? It does the same thing for me despite my laptop has all needed requirements :/
Do you know how to create nsfw content for Fanvue?
You can create NSFW images with this method.
" I don't even know what to say about that". Cmon you sound german, you prolly wanted to say "Hans get ze flamefrower"!
Hah "Get to da Choppa!"
Let's give dis beople ähr!
🚮
Good video, but really annoying voice
its over, dont do this anymore
I cant get pinokio to install fooocus. I keep getting stuck on 2/4 shell.run. Nothing is happening, it just sits there
Sorry its not working for you. I hope it's just a temporary error with the Fooocus installation and that it will be fixed shortly. You can try updating Pinokio (by installing the latest version from the website)
I ended up doing a manual install outside of pinokio, it wasnt hard and runs perfectly@@AiVOICETUTOR
jesus christ so much information this craps hard to learn
Thanks for helping fill the internet with pointless garbage for lonely losers
I get your concern but isn't that what real influencers already do. I don't mind a few real influencers losing their "job" to Ai.
Hey! I'm wondering if you know how to make the AI influencer version, but in a detailed background. I can never get that, every one I generate is with a blurry background unfortunately
Hey, I hadn't noticed that yet but thats a good observation! My first hunch would be to deactivate the style presets and see if that helps. I'll have a look myself the next time I'm on my PC
@@AiVOICETUTOR I've tried alot so far! Trying other lora's like better details, and realistic vision. Tried other models/settings/weight! The only time I can sort of get a less blurry, but still not satisfying result. Is when I put the person in inpaint, then select the background, and put the prompt as in highly detailed background. That sort of makes is 5/10% less blurry. I know it must be possible, because it can do both perfectly! Just not together haha
I've played around with a bit now and while I did get some pictures with like a 95% clear background, I have no idea how to reproduce it and it seems a bit random. I'm not sure if i'm imagining it, but it seems a bit easier to get non blurry backgrounds when the aspect ratio is sth like 1/1 or 16/9 and not portrait
@AiVOICETUTOR yes! It's very strange, it's mostly focused on a portrait. Even when you put portrait, blur, blurry etc as negative prompts! I will try out some other aspect ratio's later today thanx
Did you guys found something?