Currently, every sixel_encode does quantization, dithering, and the real encoding. This could possibly introduce 1) unnecessary computation in quantization and dithering, and 2) unstable colormap generation.
If we consider encoding a video sequence to sixel formats, then we'd like to ensure every frame/slice uses the same quantization and colormap setup.
CRef: FedeClaudi/Term.jl#94 (reply in thread)