Ai models and face consistency

[fury11]

hi, i tried fooocus and the face consistency is good but not that great for my ai models even though adding a lot of nagtive prompt to avoid eyes and hands issues ,does there any software good for consistency with quality?

 

[GOOGLEDUCK]

have you tried but with other base photos?

 

[fury11]

yes , but i have a face and i dont want a faceswap that look cartoony , i want some realistic results but with free tools

 

[MyBrocard]

If you're talking about generating a photo of a face based on parameters you set, there's Face Generator

 

[medialounge]

Check all these Deep Dream, DALL-E and RunwayML.

 

[fury11]

If you're talking about generating a photo of a face based on parameters you set, there's Face Generator

no i am talking about face swapping withe maintaining totaly realistic details of body realisticly

 

[fury11]

Check all these Deep Dream, DALL-E and RunwayML.

i tried many software they dont comply because of restrictions i found the best which is tensor art but its paid

 

[MyBrocard]

Try Remaker AI, should fit your description.

 

[fury11]

Try Remaker AI, should fit your description.

it has good results on some faces but others its trash and also the target image is somtimes off but with some twicks maybe its the best for me right now

 

[BullseyeZ]

hi, i tried fooocus and the face consistency is good but not that great for my ai models even though adding a lot of nagtive prompt to avoid eyes and hands issues ,does there any software good for consistency with quality?

Think I can help you with that.

 

[fury11]

Think I can help you with that.

can you elaborate how?

 

[RobynHood]

hi, i tried fooocus and the face consistency is good but not that great for my ai models even though adding a lot of nagtive prompt to avoid eyes and hands issues ,does there any software good for consistency with quality?

I use stable defusion. To create an initial face I find it’s best to use a prompt that is something like “who looks like ……” AI is better with people in its database than creating new people.

Note the seed number and use this in future generations. I also start all prompts with the same base

A [style of photo] photo of A beautiful brunette who looks like [name] doing [what you want to show] [lens] [lighting]

I find the less specific you are in terms of image detail the better the face quality it.

If you are using it for something important I would then run through an image to image to increase this like skin tone and imperfections

 

[fury11]

I use stable defusion. To create an initial face I find it’s best to use a prompt that is something like “who looks like ……” AI is better with people in its database than creating new people.

Note the seed number and use this in future generations. I also start all prompts with the same base

A [style of photo] photo of A beautiful brunette who looks like [name] doing [what you want to show] [lens] [lighting]

I find the less specific you are in terms of image detail the better the face quality it.

If you are using it for something important I would then run through an image to image to increase this like skin tone and imperfections

i didnt try stable defusion yet but this an intersting thing to try do you work with a lot of loras to get from a to b or just short general prompting also i need face consistency a lot can you asure that its the best for consistency after training of course

 

[BullseyeZ]

Think I can help you with that.

It's a new method on A1111.

 

[fury11]

It's a new method on A1111.

does it use a lot of gpu power?

 

[BullseyeZ]

does it use a lot of gpu power?

What GPU do you have?

 

[fury11]

i have dm you

 

[vittorianicolosi1994]

I Suggest you to use CIVITAI on A1111+ a good realistic checkpoint + LoRAs + textual inversion + additional extensions to make life easier. That's what I use, but most I've seen use SD as well
Down belle are some example



Other example

 

[fury11]

I Suggest you to use CIVITAI on A1111+ a good realistic checkpoint + LoRAs + textual inversion + additional extensions to make life easier. That's what I use, but most I've seen use SD as well
Down belle are some example

View attachment 321007

Other example
View attachment 321006

if this is your results thats a gold fucking mine with my ideas and stratgies for traffic

 

[OldWalnut]

hi, i tried fooocus and the face consistency is good but not that great for my ai models even though adding a lot of nagtive prompt to avoid eyes and hands issues ,does there any software good for consistency with quality?

How can I use Fooocus?