This is a place for experimenting with SD1.5 LoRA.
The main goal is overall enhancement rather than focusing on a single concept.
■Dora_nsfw_remember is intended to complement my test merge model, but since the training is done on NovelAI v1, they should work fine with its derivative models.
my test model: https://civarchive.com/models/1246353/sd15modellab
■nai_v2_highres is a high-resolution stabilizing DoRA for novelai_v2.
Please download the official checkpoint from the URL below.
I’ve also made a safetensors version just in case.
https://huggingface.co/NovelAI/nai-anime-v2
https://civarchive.com/models/1772131
■nai_v2_semi-real is a semi-realistic style DoRA for novelai_v2.
■I use OneTrainer for training and ComfyUI for inference.
■I will share my training settings and inference workflow as much as possible.
■If the prompt is short, the background may become simple or the style may lean toward realism.
By using the uploaded tipo_workflow, you can automatically generate longer prompts—so please give it a try!
■Sometimes saturation occurs due to overfitting and the model’s compatibility. Adjusting DoRA , prompt weight strength, or reviewing the cfg can help improve this.
■There’s also a high-resolution inference workflow using kohya_deep_shrink.
It expands composition and removes the need for high_res_fix.
1152px offers a good balance of quality, stability, and speed, while 1536px is more dynamic and detailed.
By the way, this Dora was created to give SD1.5 a level of concept understanding comparable to my PixArt-Sigma anime fine-tune.
SD1.5 with Dora applied will likely be the most compatible refiner—it's ideal for i2i tasks.
my pixart-sigma finetune.
https://civarchive.com/models/505948/pixart-sigma-1024px512px-animetune
Description
Training Data (46.28 MB):It contains the LoRA from epoch 41. The training is weaker, but it is less likely to break down.+comfyui workflow,onetrainer data
●As an experiment, I trained a Dora using OneTrainer with a dataset of about 17,000 images.
●The main focus was strengthening recognition of hard-to-detect NSFW tags. Around 3,000 images were randomly selected for general NSFW reinforcement, while the rest focused on specific tags with dedicated datasets.
Though quality isn't guaranteed, it may help guide concepts.Not all concepts have been perfectly verified, so there may be some undertrained tags.
Since the scale is small, there are limitations. I believe it is weaker than a 1 concept LoRA, so please consider it as just a subtle adjustment.
● Training was done on NovelAI_v1, so it should work with its derivatives.However, some models suffer from concept forgetting due to overfitting on specific concepts or styles. What works in NovelAI may be weaker elsewhere. Increasing the LoRA weight above 1 or strengthening the tags might help.There are few issues up to a strength of 2. Conversely, if artifacts increase, try lowering them.
●Be careful with negative prompts.
Using terms like "worst quality, low quality:1.4" can easily produce high-quality images, but they may limit diversity.
If special tags like "slime girl" don’t seem to have an effect, the strength of these negative prompts might be too high.
●Training with NovelAI after a long time was easy and enjoyable. Since the results were good, I might try using the 400,000 dataset, focusing on general tags that are harder to recognize.
●This Dora does not have trigger tags. The goal is general concept enhancement, not restricting the original model.
This is not a major modification; it's just guiding concepts that the model already knows but have been buried and are not functioning properly.
●I aimed to evenly learn as many concepts as possible. On its own, the effect is weak, so using multiple related tags may be necessary.
If the character has long green hair, it's best to include those tags. The same applies to NSFW tags.
●Additionally, for minor concepts, there may be a lack of high-quality data, so good results might not be guaranteed.Also, since I enjoy visual novels, my style may lean towards that.
●There may be occurrences of "censored, mosaic censoring, bar censor" but that is because these are included in my dataset. I also include uncensored images, but many images have censorship. You can enjoy them as expressions or spice, or try adding them to negative prompts. It might also be a good idea to add "uncensored" to the prompt, although it might not always resolve the issue.
●I don't train many LoRAs, so I would appreciate any feedback or ideas you have about this LoRA. Depending on the situation, I may use it as a reference to make improvements.
●I'll also share my OneTrainer settings for reference.
This is U-Net-only training. Training with Clip Skip 2. For OneTrainer, it's set to 1.I'll also share my ComfyUI workflow just in case.
■Focus train tag: There are also tags that are undertrained.
against glass
corruption
doggstyle
body writing
footjob
futanari, 1futa, 2futa etc...
paizuri
penis
uncensored
wariza
slime girl
petrification
tentacle clothes
■chara:
aragaki ayase
run elsie jewelria
may (pokemon)
There may be other tags that will be strengthened as well.