SUPERMACHINE

Product details

Q: Is there a way to create our own consistent characters?

I am looking for something where I can design a character through trial and error, then essentially save that character, then put it in different situations.

Also, I tried creating prompts with real faces (groups of people) and they all came out like they had been shot in the face. Can you improve face generation?

gkcPLUSNov 26, 2023
Founder Team
Rexxar

Rexxar

May 14, 2024

A: Hey there,

It is totally possible to make real faces. Check out some of these galleries
- https://supermachine.art/smv2-modellanding/supermachine-photo
- https://supermachine.art/smv2-modellanding/supermachine-reality
- https://supermachine.art/smv2-modellanding/aof-model

We also have a bunch of consistent characters for users to use which can be seen in these galleries
- https://supermachine.art/smv2-char-landing/clive-smith
- https://supermachine.art/smv2-char-landing/jessica-langridge
- https://supermachine.art/smv2-char-landing/maggie-smith

I think you'll find that the images above are not from gun wound patients. The complexities of this technology mean that prompting is so important. You can see the prompts used in the images above by just clicking on them, you can also use tools in the product like the image 2 prompt tool or the Prompt Book to see successful prompts and help you with your image generation journey.

At the moment, we don't have a way for users to create their own characters but we are exploring ways to help users with this. Through our own consistent character creation, we believe that we have created a method which is far superior in quality to anything else that we have seen and by just using a single image, you won't get that quality which is some of the challenges involved here. We're trying to automate the process we have and are making strides in that. At the moment as LORA introduction and controlnet is the priority for us to get done by the end of this year.

Thanks!

Share
Helpful?
Log in to join the conversation
gkcgkcPLUS
Verified Purchaser badge

Verified purchaser

Posted: Nov 26, 2023

Hey!

Thanks for taking the time to write this out.

Those are really good faces. Do you think the problem is I had 5 or 6 faces in one image? I.e., a group of people. For example, my prompt included "Group of women dressed in late 1800s attire", and it got the group of women and attire right, but they all have holes in their faces or distorted faces. Seems like it should just create non-distorted faces.

With the consistent characters: am I right in understanding that LORA and ControlNet could enable this?

Sorry, could other basics:

1. Will AppSumo users get all future features?
2. Do we own the copyright to our images, and can use them commercially?
3. Can anyone else see our images we make. I used another platform that made all our images public by default.

Thanks again!

gkcgkcPLUS
Verified Purchaser badge

Verified purchaser

Posted: Nov 26, 2023

I'm sorry, one other thing. Can we do inpainting, or generally addto an image. Like if I upload an image of a character, can SuperMachine change the background, but leave the character as it was?

Founder
Posted: Nov 26, 2023

Multiple people in an image can cause an issue. It can generally work well up to 2 or 3 but more than that can cause issues.

In terms of the Appsumo questions;
1. Yes, will always be able to use everything in the product.
2. You can use the images commercially. We don't claim any copyright over anything you make.
3. Everything in Supermachine is private by default. You can make albums public if you want to share images with friends / co-workers but privacy is the default in Supermachine.

Thanks