I migrated this project to <a target="_blank" rel="ugc" href="https://civarchive.com/models/2846802/fwaun-tune">Fwaun Tune</a>.<a target="_blank" rel="ugc" href="https://civarchive.com/models/2846802/fwaun-tune">Fwaun Tune</a> に移行しました。A fine-tuned model specializing in flat color, no lineart, slightly chibi character art.フラットカラー、主線なし、デフォルメ（ちびというほどじゃない）の画風を学習させたファインチューンモデルです。<h2 id="how-this-model-was-made-fwaun-tools">How this model was made / このモデルの作り方 — <a target="_blank" rel="ugc" href="https://github.com/fwaunstp/fwaun-tools">fwaun-tools</a></h2>This model was built with an open-source pipeline I develop, <code>fwaun-tools</code>. Sharing the workflow for transparency:<ol><li>Dataset prep — tags and captions are curated with <code>fwaun-tools</code> (manual editing + WD14 tagger + captioning with OpenAI API + Danbooru tags in one surface).</li><li>Training — the base model is trained with <a target="_blank" rel="ugc" href="https://github.com/kohya-ss/sd-scripts">sd-scripts</a> on that dataset.</li><li>Diff extraction &amp; merge — <code>fwaun-tools</code> extracts the full fine-tune delta (<code>tuned − base</code>) and transfers it onto a Turbo checkpoint with <code>merge-diff</code>.</li><li>Quantization — the merged checkpoint is quantized to INT8 (+ConvRot) with <code>fwaun-tools</code>.</li></ol><code>fwaun-tools</code> 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 <code>fwaun-tools</code> 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:<ol><li>データセット準備 — タグとキャプションを <code>fwaun-tools</code> 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)。</li><li>学習 — そのデータセットで <a target="_blank" rel="ugc" href="https://github.com/kohya-ss/sd-scripts">sd-scripts</a> を使ってベースモデルを学習。</li><li>差分抽出・マージ — <code>fwaun-tools</code> の <code>merge-diff</code> でフルファインチューンの差分(<code>tuned − base</code>)を抽出し、Turbo チェックポイントへ転写。</li><li>量子化 — マージ後のチェックポイントを <code>fwaun-tools</code> で INT8(+ConvRot)に量子化。</li></ol><code>fwaun-tools</code> 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。<h1 id="v4.0">v4.0</h1><ul><li>Change trigger words.</li><li>Update dataset to increase the character's head-to-body ratio</li></ul><h1 id="v3.0">v3.0</h1>To help body proportions, I’ve added super deformed, chibi-style illustrations to the training data.You can decrease the body-to-head ratio using the “chibi, super deformed” tags and you can also use "realistic" tag to increase it.Trigger word<pre><code>flat color, no lineart, deformed</code></pre>Optional tags<pre><code>chibi, super deformed</code></pre><pre><code>realistic</code></pre><h1 id="v2.0">v2.0</h1>To help adjust color tones and body proportions, I’ve added non-deformed flat color no lineart illustrations to the training data.You can increase the body-to-head ratio using the “realistic” tag. If you don't want to increase it, set it to a negative value.Trigger word<pre><code>flat color, no lineart, deformed</code></pre>Negative word<pre><code>chibi, realistic</code></pre><h1 id="initial-release">Initial release</h1> Note: This is my first fine-tuned model. While I'm pleased with this results, the quality may vary compared to established style LoRAs. I appreciate your patience and feedback as I continue to refine this approach. Your input helps me improve future versions! This model excels at creating characters with a distinctive light deformation style - more expressive than realistic, yet more refined than super-deformed (SD/chibi). Perfect for characters ranging from 3.5 to 7 heads tall, capturing that sweet spot between youthful charm and mature appeal.Key characteristics:Clean flat colors (no lineart required)Soft, natural deformation - ideal for stylized character designWorks beautifully with other character LoRAs to apply different character traits while keeping this style intactStable on deformed proportions that danbooru tags struggle to captureRecommended usage: Combine with character LoRAs for maximum flexibility. The model's style strength is intentionally calibrated to enhance, not overwhelm, other adapters. <h2 id="training-details">Training Details</h2>For transparency and to help with troubleshooting, here's the training command I used with sd-scripts:<pre><code>accelerate launch --num_cpu_threads_per_process 1 anima_train.py --pretrained_model_name_or_path=/path/to/animaOfficial_preview3Base.safetensors \
	--qwen3=/path/to/qwen_3_06b_base.safetensors \
	--vae=/path/to/qwen_image_vae.safetensors \
	--dataset_config=/path/to/dataset_config_anima.toml \
	--output_dir=/path/to/outputu \
	--output_name=fwaunstp_ckpt_anima_preview3_v1 \
	--save_model_as=safetensors \
	--learning_rate=1e-6 \
	--optimizer_type=AdamW8bit \
	--lr_scheduler=constant \
	--timestep_sampling=sigmoid \
	--discrete_flow_shift=1.0 \
	--max_train_epochs=8 \
	--save_every_n_epochs=1 \
	--mixed_precision=bf16 \
	--gradient_checkpointing \
	--cache_latents \
	--vae_chunk_size=64 \
	--vae_disable_cache \
	--save_state \
	--split_attn \
	--attn_mode=torch</code></pre>This model was trained using f32 precision (not BF16, which some other Anima Preview 3 models use). If you notice any issues or compatibility problems, please let me know—I'd appreciate feedback on the training parameters or the checkpoint format.

Fwaun Anima

fwaun_anima_v30_turbo_int8_convrot.safetensors

fwaun_anima_v30_turbo.safetensors

v3.0_turbo

Files with SHA256: 3b4f630a6b11ef9c3335b3456c6990abb192713127377370a293262f4528bd82