AI Core
Defines the AI Core to use. Different cores are trained with different data sets, depending on your style image, you will want to select the core that gives the best results. More AI Cores will come in the next updates.
Core 1 – Generic
Basic training, not very detailed but works great with most cases.
Core 2 – Painting Detailed
Gives more detailed result, better with a style that looks like a painting.
Here’s a comparison between two cores, using the same
source image and
style:
Style
Select another layer as the style image.
This other layer shouldn’t be too large. 400×400 px often gives the best results.
Tip: select a pre-comp as the style source and play with the scale of the image inside this comp to change the look of the result. Note that you can also animate the style inside the pre-comp, or change it every frame to do an hand-drawn animation style.
Ratio
Defines how much of the original image goes into blending. 100% will give the strongest style.
Keep Original Colors
Check this option to preserve your source image’s color. The style features and details will be applied but not the styles predetermined color palette.
Reduce Flickering
3 tools are proposed to reduce frame-to-frame flickering when applying Transfusion to a video or an image sequence:
- Reduce with blur:
Will add a small amount of blur to the source image, resulting in a smoother and less noisy input, this is likely to reduce or remove noise and grain in video.
- Reduce with noise:
Will add a subtle static noise pattern over the image. This noise will help the algorithm to stick the style features at the same place on every frame, especially in large and uniform areas.
- Start from last result:
When activated, Transfusion uses the last rendered frame as the input frame, resulting in a much smoother output when working with video.
The Mix slider defines how much of the last render is used. Typically, a lower value (between 5 and 20%) is enough to help the algorithm produce a stable result and still keep almost all of the animation.
