A Signature mode is a Composition Control tool which combines multiple base modes to create a blended and balanced multipurpose tool. Signature modes are typically designed to support a particular use case or style.
How are Signature Modes different from Base Modes?
Base modes, such as Pose or Structure, use one pre-processor and one composition control model during their image creation process.
Alternatively, Signature Modes consist of two or more base modes that Scenario has explored for a particular style, subject or use case. Signature modes use multiple base modes together and highlight several distinct details of an image’s composition. This allows for more targeted and complex interpretations of reference images. Below is an overview of what elements are combined.
City Mode
City mode combines Edges, Depth, and Segmentation modes. This means that it is paying attention to perfectly straight lines, the area in space that is taken up, as well as the depth of field in an image.
For example, for this reference image:
Three mode maps are detected:
.png?width=688&height=313&name=grid%20(8).png)
using our Asset Diffusion model with the prompt
a staircase in a post apocalyptic survival rpg, 3d render style, dark, spooky, haunted
we can see the following outputs:
_a%20staircase%20in%20a%20post%20apocalyptic%20survival%20rpg%2c%203d%20render%20style%2c%20dark%2c%20spooky%2c%20haunted_image-2_1685561353.png?width=688&height=688&name=Asset%20Diffusion%20(Beta)_a%20staircase%20in%20a%20post%20apocalyptic%20survival%20rpg%2c%203d%20render%20style%2c%20dark%2c%20spooky%2c%20haunted_image-2_1685561353.png)
Character Mode
Character mode combines depth and pose mode. It is designed to both detect the pose as well as the general features of an input image, while still providing space for the prompt and generator style to show.
For example using this reference image:
Generates these mode maps, which are compared:
.png?width=688&height=313&name=grid%20(9).png)
Combined and balanced, they produce this output using Asset Diffusion and the prompt:
a super high tech alien creature, bionic, cyborg, 3d render style
_a%20super%20high%20tech%20alien%20creature%2c%20bionic%2c%20cyborg%2c%203d%20render%20style_image-2_1685561584.png?width=688&height=688&name=Asset%20Diffusion%20(Beta)_a%20super%20high%20tech%20alien%20creature%2c%20bionic%2c%20cyborg%2c%203d%20render%20style_image-2_1685561584.png)
Landscape Mode
Landscape mode combines Structure and Segmentation modes. This preserves lots of nuanced details, which is balanced against overall blocks of space. This is particularly useful for creating environments in multiple seasons, and using real references for game backdrops.
Using this reference image:
That in turn generates the following pre-processed images which are interpreted within the AI:
.png?width=688&height=313&name=grid%20(12).png)
We can create a environment or backdrop which resembles and entirely different style using Asset diffusion and the prompt:
a strange alien world
using the 80s anime public generator:
Interior Mode
Interior mode is a combination of edges mode and segmentation mode. It is designed to clearly outline the manmade, as well as natural aspects, within an image of the interior or a building or vehicle.
Interior mode can take an image such as this:
And create the following maps:
Which can be powerfully used to reinvent and reskin the final mode map, in this case using Asset Diffusion and the prompt:
happy candy room
_a%20happy%20candy%20room_image-0_1685564045.png?width=688&height=688&name=Asset%20Diffusion%20(Beta)_a%20happy%20candy%20room_image-0_1685564045.png)
Conclusion
Signature maps are a great introduction to the power of blending mode maps, which can also be achieved using the Scenario API. We will continue to release relevant mode maps, so stay tuned!