<div data-type="blurb" data-id="304">👉<a target="_blank" rel="ugc" href="https://www.patreon.com/hshiKoi/posts/174642922">【488 Pictures】hs krea2_turbo Sample Images (Workflow Included)</a>👉 <a target="_blank" rel="ugc" href="https://www.patreon.com/hshiKoi/posts/jian-xin-zhui-yi-174665305">【81 Pictures】 舞い散る雪にして、優しき鞘。</a></div>An all-in-one ComfyUI workflow for Anima (2B Anime Foundation Model), offering calibrated sweet-spot generation, 3 distinct upscaling solutions, and automated batch prompt testing.✨ Key Features- Calibrated Anime Baseline: Tuned for Anima's official sweet-spot resolutions (1MP / 2MP, multiple = 32) for stable anatomy and clean lines.- Dual-Mode Sampling: - Turbo Mode: 12 steps, CFG 1.0, er_sde + simple (with Turbo LoRA). - Base High-Fidelity Mode: 28–30 steps, CFG 4.0–4.5, er_sde + simple (LoRA off).- 3-Way Switchable Upscale &amp; Refine: 1. Anima Dual Upscale (Main): Native 1.5× latent pass with zero extra VRAM overhead. 2. SeedVR2 (4K) Upscaler: Dedicated 4K ultra-resolution enhancement fed straight from the base pass. 3. Flux2-Klein Adaptive Refine: 1.5× redraw with automatic prompt syncing and ReferenceLatent conditioning.- Automated Batch Testing: Built-in line-by-line prompt reader (up to 100 prompts) for effortless batch evaluation.### 💡 Quick Guide- Generate base anime images in [Hs Anima Txt2Img].- Select your preferred upscaler in [Fast Groups Muter] (enable one at a time for optimal VRAM efficiency).- (Optional) Press <code>Ctrl+B</code> on the batch loader to auto-test prompts from a <code>.txt</code> file.<hr />🌱 Multi-LoRA 4K Workflow<ul><li>Multi-LoRA support</li><li>Flux2Klein enhancement</li><li>SeedVR2 4K upscaling</li><li>High-quality final output</li></ul><hr />Anima + LoRA + Face Repair + HD Upscale Workflow<ul><li>Anime character face repair</li><li>Face detail enhancement</li><li>High-quality upscaling</li></ul>Anima+LoRA+面部修復+高清放大工作流程<ul><li>動漫角色修臉</li><li>面部細節增強</li><li>高畫質放大</li></ul><hr />作者資訊 Hshi微信公眾號：画不出明天QQ群：978336169微信群：hshi2026加入時請註明來意。

Hshi💗Anima 4K Master Workflow (Txt2Img & Multi-Upscale)

<div data-type="blurb" data-id="304">👉<a target="_blank" rel="ugc" href="https://www.patreon.com/hshiKoi/posts/174642922">【488 Pictures】hs krea2_turbo Sample Images (Workflow Included)</a>👉 <a target="_blank" rel="ugc" href="https://www.patreon.com/hshiKoi/posts/jian-xin-zhui-yi-174665305">【81 Pictures】 舞い散る雪にして、優しき鞘。</a></div>Major update bringing a restructured 2-pass pipeline, modular upscaling, and batch testing support.What's New in v2.0- Sweet-Spot Baseline: Pre-calibrated 1MP / 2MP sweet-spot aspect ratios via <code>ResolutionSelector</code> to ensure stable anatomy and line-art consistency.- Upgraded 3-Way Upscale Suite: 1. Anima Dual Upscale: Native 1.5× latent pass with zero extra VRAM overhead. 2. SeedVR2 (4K) Upscale: Direct high-resolution detail restoration pipeline. 3. Flux2-Klein Refine: Adaptive 1.5× redraw with automated prompt synchronization.- Automated Batch Testing: Added <code>Text Load Line From File</code> node (Ctrl+B to enable) for bulk prompt testing (up to 100 prompts).- Dual Sampling Presets: Optimized parameters for Turbo LoRA (12 steps, CFG 1.0) and Native Base mode (28–30 steps, CFG 4.0–4.5).

Files with SHA256: 0d42d441cc3e4c859f1588cee449e401550aed994c6f15d14074fd5f12b06efe