I migrated this project to Fwaun Tune.
Fwaun Tune に移行しました。
A fine-tuned model specializing in flat color, no lineart, deformed character art.
フラットカラー・主線なし、(ちびというほどではない)デフォルメ画風を学習させたファインチューンモデルです。
How this model was made / このモデルの作り方 — fwaun-tools
This model was built with an open-source pipeline I develop, fwaun-tools. Sharing the workflow for transparency:
Dataset prep — tags and captions are curated with
fwaun-tools(manual editing + WD14 tagger + captioning with OpenAI API + Danbooru tags in one surface).Training — the base model is trained with musubi-tuner on that dataset.
Diff extraction & merge —
fwaun-toolsextracts the full fine-tune delta (tuned − base) and transfers it onto a Turbo checkpoint withmerge-diff.Quantization — the merged checkpoint is quantized to INT8 (+ConvRot) with
fwaun-tools.
fwaun-tools is a personal project — not heavily tested, so bugs are likely — but issues and pull requests are welcome.
このモデルは、自作のオープンソースツール fwaun-tools を使ったパイプラインで制作しています。透明性のためワークフローを公開します:
データセット準備 — タグとキャプションを
fwaun-toolsで整備(手動編集+WD14タガー+Open AI API 互換キャプショナー+Danbooruタグを1画面に統合)。学習 — そのデータセットで musubi-tuner を使ってベースモデルを学習。
差分抽出・マージ —
fwaun-toolsのmerge-diffでフルファインチューンの差分(tuned − base)を抽出し、Turbo チェックポイントへ転写。量子化 — マージ後のチェックポイントを
fwaun-toolsで INT8(+ConvRot)に量子化。
fwaun-tools は個人開発のツールです。テストは十分ではなく不具合の可能性がありますが、イシューやプルリクエストは歓迎します。
Description
Change trigger words
Update dataset to increase the character's head-to-body ratio
FAQ
Comments (3)
Just FYI to anyone who's curious, this model seems entirely censored
That’s because this model is based on Krea 2 Turbo, and Krea 2 Turbo is censored. Since this model itself hasn’t been trained with additional censoring (Actually, the trained dataset contains some uncensored images) , it’s possible to combine it with a different uncensoring LoRA. Just for your information.
@fwaunstp733 Thanks for the explanation.


