1) Dangerous techniques and tools
Off the top of my head:
Color tools
All the "Color" tools (Curves, Levels, Contrast) entail some color loss (which translates into visible banding in the bad cases)(this not Gimp's fault, this is math). This can be mitigated by:
- using high precision images in Gimp 2.10
- avoiding accumulating changes. Ideally everything can be done with one single application of Curves
Transform tools
All the "Transform" tools (except the trivial flips and N*90° rotations) require pixel interpolation that adds some amount blur. This can be mitigated by:
- doing everything in one single operation (see the Unified transform tool in Gimp 2.10)
- working with paths and rendering the path (stroke/fill) once the transform has been applied (especially if working with text)
Selection tools
Fuzzy select and Color select do "binary" selection (a pixel is fully selected or not at all). Used alone without care, they are the biggest enemies of smooth edges in CGI because they remove the anti-aliasing pixels. See Colors>Color-to-alpha
and Color erase
modes in paint tools and layers.
2) XCF format
XCF is a storage format, but when it memory the image is the same if loaded from XCF or PNG or GIF... (though GIF is color-indexed, and editing GIF should not be attempted by untrained personnel). XCF helps maintaining quality because
- It is a lossless format, no pixels are altered when saving
- If you use high-precision images in 2.10, the XCF keeps the high precision data, there is no conversion to 8-bit as you would have if saving to JPG or PNG (some more recent format also support 16-bit data)
- It saves everything (layers, channels, paths...) so you don't have to redo things (that would rarely be redone with as much care...)
3) Blend mode
You use a blend mode because you need it. This said, in low-precision images piling layers with various blend modes can lead to round-off errors and artifacts such as banding. Again, using high-precision images reduces that risk.