Building and using References

In addition to Characters, References also help with static definitions of objects and poses that helps with scene building.

Much as Character building provide consistency across images and videos produced, so too does the further definition of what is considered "References".

References can be categorised into four types:
1) My Objects - item or things that can be placed or interacted with.
2) My Locations - Scenes where you character or image will be located in.
3) My Poses - Poses that you want your character to copy or emulate.
4) My Outfits - Clothing that you want your characters to wear.

You start by clicking on References on the on the left hand side menu. On the main screen, you can see other Objects already created and their "@" handle which you can use to call these objects into prompts.

Click "Create New"

Let's create an outfit

Upload a photo and you'll need to crop it as much as possible before uploading otherwise it wont render properly.

I've found a hat from the H&M website that I want to use, so I snipped the image, cropped all the extra details around it and then uploaded this under Outfit and named it "Hat". The system will allocate your outfit a random handle name which you can call up in prompts.

Click Save.

Building the other types are exactly the same and you'll want to do as many of these as possible to a) reduce prompt strain and also to b) ensure consistency

Now that we have an outfit and a character, we want to combine the two using a prompt. Once again, I use Mango 3 since I think its superior, but you can experiment with the others too.

Applying the outfit

The result works to some extent. Demon Bear is wearing the hat, but was this what you imagined it to be? That difference is because I didn't properly illustrate the background, location, zoom, angle, etc. AI will use it's own initiative to provide a picture, but unless specified, it will just do what it wants.

Now that we have dressed him in an outfit, lets put Demon Bear into a actual familiar movie location. I've already created a location object and will apply it to the prompt.

Now add a scene

The result does what it says, but did turn Demon Bear white. Not sure why it did that.

I did end up using the prompt help as well to properly articulate the prompt. But it messed it up completely, so I had to rewrite the prompt in "real speech".