I'm bringing this new training done with some screenshots of pc-98 old japanese games, this is intended to bring old anime aesthetic and pixelart.
Description
FAQ
Comments (8)
This brings me back. It's a shame your examples don't have prompts in them, but so it goes. Thank you for your hard work in any case.
Save video -> drop on a comfyui workflow
@souryuusen I don't use comfyui (I use forge, even though I know it's worse, I haven't found the time to figure out switching), but even if I did, it's pretty tedious to have to load it into comfyui. There are nodes that can help attach prompts in ways that civit (and other tools) can see, even if your normal prompt setup doesn't work for this.
It's nice to make the prompt visible, at least for the images used to demonstrate a LoRA and such (even if you don't want to do it other times), since it helps show how the creator of the LoRA expects it to be used (or at least how they use it themselves). I suppose a comfyui workflow does this too, but it's tool-specific and also more annoying to pull in, I think (also there's no indication it has anything like this on the civitai website).
Anyway, I still like the LoRA even if you don't do these things, but it would be nice.
@notone sorry for that, i usually use a LLM to create the prompts using a reference image as base, so the prompts are dynamic and because i am testing the lora consistency i create several videos with the same template and those are not in the metadata of the comfyui exported video, i ask to the LLM to create the image/video prompt with pc-98 old visual novel aesthetic and pixelart, however if you include in your prompt words like pc-98, pixelart, visual novel usually is enough to get the effect of the lora, also increasing the strength till 2 it maximizes the effect, the aesthetic also is different from one model to other, personally i like more the Anima outputs.
@notone I'm not sure which lora the comment refers to, but I posted in the gallery for the minimax h3 version. I included my prompt. Hopefully it helps.
It was quite a challenge for reference to video, probably much easier for text to image/video. Simply "pixel-art, PC-98, anime-style" wasn't enough haha.
Great work!
Big PC98 fan here, thanks a lot!
Always love loras in this style