Our aim with this checkpoint, & other Colorize Truer/+ LoRAs, is to help restore Kontext Dev towards the far more impressive photo colorization capabilities of Kontext Pro/Max, whilst further reinforcing task-specific reliability & realism at the standards of professional manual colorization.
This checkpoint was merged from several of our trained variants + an additional colorization LoRA that takes reference inputs (via typical control input channels). Alas, this color reference feature is not yet likely to function reliably/predictably.
The merge was done using the sd-scripts
Flux LoRA merge script.
The constituent models were trained with
AI Toolkit by Ostris over 500 grayscale/color pairings of real photos.
The source photos were sourced from public domain, and represent a broad scope of geographical, cultural, aesthetic, historical (from late 1830s thru today), & other paradigms/contexts.
No trigger words neccessary.
With that said, with some inputs, it may be neccessary to specify colorize or colorize and upscale this photo. Leave all else the same. as the instruction prompt.
One may also prompt in suggestions/corrections/directions, as far as specific choices of color.
To "modernize" outputs, one may try: Colorize and upscale the photo. Restore crisp detailing. Rich tonality. Make indistinguishable from top quality UHD 50MP+ or 12k+ pro 2025 DSLR camera colorized_02.CR2 photo. restore blurred details. Retain facial features with great fidelity source. Add fine detailing only where missing. Render real life like tones. Leave unchanged all else about photo., etc...
Make sure to match-up input/output dimensions, primarily to ensure fidelity of details/identity.
With starkly differing dimensions, the model might also get confused in regards to its instruction.
