Category: Uncategorized

What a term! We explored so much about text to image AIs, however, we must not ignore the environmental impact of AI training. As we discussed in class and attempted during our assignment, LORA training requires a lot of energy even with advanced GPU. So I’d like to present a mini exploration on the power…

I think Kohya_ss really hates non Nvidia GPUs. Tried with a lot of different configurations, most of them resulted in out of memory error. (I am on a 16GB AMD Radeon GPU!). Even with successful runs (extremely slow and loud from coil whines) results were undesirable. I really wish there could be a Kohya_ss alternative…

I could be completely wrong here… But recently I switched to an AMD RX6900XT 16GB from an Intel Arc A770 16GB. The computing API was changed from Intel IPEX to AMD ROCm, and something weird happened. Again, I could be completely wrong here but I found that my ComfyUI suddenly started to output higher quality…

I generated a bunch of Chinese liminal space images for this assignment Prompt in danbooru style without weight: liminal_space, 2000s_shenzhen, huaqiangbei, street_in_middle, nostalgic_vibes, clean_street, parked_bikes, old_apartment_blocks, white ceramic_wall, blue_window, local_shops, faded_paint, street_signs, parked_cars, cable_hanging_between_buildings, street_lights, retro_signs, soft_lighting, summer_evening, childhood_memories, melancholic_atmosphere, quiet, 2000s_architecture, no_people, lost_time_aesthetic Text to Image (some with Lora) These images were generated in…

Since I am using Comfy instead of a1111, the inpainting process is a bit different. Instead of painting directly in WebUI, I need to erase parts that I need to inpaint with Pinta and upload the erased picture. The result can be quite interesting: And then there is… It really is a model that’s more…

Unfortunately I lost my first week’s prompt text, and ComfyUI does not embed any metadata into images, I had to use a different set of prompt. Different checkpoints can either have different features (e.g. lora compatibility) or art styles. The following pictures shows the same prompt but under different checkpoints :

The goal of me taking this class is to learn how to properly write prompts, explore alternative prompt generation methods, and explore AI generated audio and visuals. I also wish to learn how those AI work at a lower level.

Welcome to Moene’s Sites. This is your first post. Edit or delete it, then start writing!