Tools, news and tutorials for graphics software and hardware. Also AI!
The free open-source G’Mic 4.0 is now available. The GMIC-8bf filter for Photoshop is the one most people will want, and specifically GmicPlugin_x64.zip for Windows. Has an interesting new take on the ‘transfer colours’ filter… New filter ‘Colors / Transfer Color [Multi]’ proposes several different methods to transfer colors from a style image onto a... Continue reading »
ComfyUI-PixelDriftFix… a ComfyUI node that really does fix pixel drift (at last!) by using some clever behind-the-scenes ‘homography matrices’ maths. In other words: what goes in now matches what comes out, in terms of a really good 1:1 overlay when brought back into Photoshop. One can see the potential for local colorisation, via a Color... Continue reading »
When loading images from eBay in Photoshop, such as postcards at 1600px wide, it’s annoying that the right-hand side of the images is hidden under the UI’s side-panels. When what you want is this (below). Now we have the whole image in view, and the side-panels on the side… Double-clicking the top-bar on the panels... Continue reading »
* Comic-bubble Eraser, a new free Action for Photoshop plus three scripts to enable the Action to work. Wholly offline, no AI or Python. No Adobe log-ins required for ‘extra downloads’ and suchlike. Gives a clean result on standard-layout Eurocomics, but is not designed for manga. For scanlations, use it in tandem with the free... Continue reading »
For the free ComfyUI, the free comfyui-remove-text custom node. This removes text locally. With no OCR, no masking, and no huge LLM needed. Includes and calls its own tiny 4Mb DBnet AI model, for processing the positions of the words, which comes along with the custom node. Yes, you read that right, 4Mb, not 4Gb.... Continue reading »
Oh dear, despite their previous permissive Creative Commons status, you can no longer use David Revoy’s Pepper & Carrot (aka Pepper&Carrot) comic-strips with AIs/LLMs or anything using them. This includes just demo-ing the capabilities of LLMs. The terms have been changed for all of “the Pepper&Carrot web comic and related projects”. From Pepper & Carrot... Continue reading »
A demo of having the new Qwen3.5 4B local AI make a quick translation of a comic-book page. Qwen3.5 (2.6GB in size, as a .GGUF) was run locally with its Vision mode in Jan.ai on Windows. Jan.ai runs the latest Llama framework. Qwen3.5 is hugely multilingual and can speedily OCR/translate. RESULT: PROMPT: Create a HTML... Continue reading »
The image-generating Z-Image Turbo has been out since mid December, and I’ve collected a judicious selection of style LoRAs for it. I’ve finally found time to put them all to a simple group-test with its fast Nunchaku variant model, released at the very end of December 2025. Which LoRAs would straightaway affect the prompted image,... Continue reading »
The amazing Z-Image Turbo image-generator AI model just became even faster, with the release of a Nunchaku version. Which is kind of ‘Turbo-Turbo’. For me, it’s very significantly faster, even for a relatively large image. Here we see a 1280 x 768px image arriving in just over seven seconds. Even with a style LoRA in... Continue reading »
A few more Z-Image Turbo demo pictures (hard to call them just “images” now), hopefully proving that it’s about more than photographs of busty waifus… Better than SDXL? Well, that depends. No negative prompting. But the prompt adherence is often so good, they’re often not needed. But, more complex prompts are needed. Even at 960px... Continue reading »