← sources

Current best models for image gen - beginner questions :-)

source: reddit (reddit:comfyui)published 2026-10-19 06:19ingested 2026-10-20 02:39

retained text

Hi everyone, So I've been using online platforms for image and video generation. I've also done some limited training of Loras on RunPod.

I've now got a dedicated machine to do this stuff locally 5090, 128gb ram . I am starting to download the models for image and video gen, but things move pretty fast and wanted some opinions for which models work best from a lora training perspective, image generation and video generation.

My shortlist is, for image gen and training for character and style Lora's; Flux.2-klein, Krea 2 FP8  (16:9 - 1080) although I could use GPT2 for plates 16:9 at 1080.

For local video I don't see anything better than Minimax H3 right now, but could be off base there? Are there good alternatives?

Sorry for the noob questions. :-) and thanks for any responses

document fetched 2026-10-20 00:41 · canonical https://reddit.com/r/comfyui/comments/1x9pa9h/current_best_models_for_image_gen_beginner