AnimaForgeFree, local, one-click LoRA trainer built only for Anima. Point it at a folder. Name it. Forge.Training a LoRA shouldn't mean an evening lost in Kohya's wall of knobs. AnimaForge is a free Windows app that does the whole run from one screen: pick your image folder, name the LoRA, hit Train. It auto-captions (WD14 tags + JoyCaption descriptions), calculates sane step counts for your dataset size, and trains with settings hand-tuned for one model — Anima. That focus is the whole point.You watch it learn. Preview images render every epoch while training runs. Flip on Compare epochs to see them side by side, spot the earliest epoch that already nails your character, and stop right there — no over-training guesswork, no re-runs. Stopped too soon or crashed? Resume from the last checkpoint.It has opinions, so you don't need any. Pick a preset — Person, Object / Concept, or Style — and the subject handling, optimizer, network size, and step budget are set. Power users can save their own presets, switch optimizers (Prodigy auto-LR or classic AdamW8bit), and preview the exact training config before launch. Guard rails catch the classics before they burn a run: empty captions (one-click fix), low VRAM, even duplicate images hiding in your dataset.It finishes the job. One click copies the trained LoRA into your Forge or ComfyUI models folder — with the trigger word baked into the filename, so anyone you share it with knows exactly what to type. A batch queue runs multiple LoRAs unattended and survives a restart.Runs on your machine. Your images, captions, and LoRAs never leave your computer. No subscriptions, no queues, no uploads.WHAT'S NEW — v1.9• Training presets: one-click intents (Person / Object / Style) + save your own in Setup• Compare epochs: side-by-side previews per epoch — stop at the earliest best• Deliver to ComfyUI + trigger word in the delivered filename• Duplicate catcher: same-name images flagged in the gallery, one-click delete• Cleaner, more readable UI (proper CJK support), accurate speed/ETA dials, RTX 50-series readyFull history: github.com/SillySilk/AnimaForgeWHAT YOU'LL NEED• Windows + NVIDIA GPU (CUDA) — includes RTX 50-series. No CPU/AMD path.• ~16 GB VRAM comfortable; low-VRAM mode reaches smaller cards (8 GB practical floor).• The three Anima files (Setup auto-detects them): DiT checkpoint, Qwen3 text encoder, Qwen-Image VAE — from Hugging Face circlestone-labs/Anima.GET IT → github.com/SillySilk/AnimaForgeinstall.bat (one-time — builds a self-contained .venv with the full stack), then launch.bat.100% free and open source (MIT). Built for the Anima community. Forge something great.------------------------------ ② VERSION DESCRIPTION (v1.9) ------------------------------v1.9 — presets, epoch comparison &amp; your feedback• Training presets — one-click intents (Person / Object / Style), plus save your own in Setup for bigger configurations• Compare epochs — side-by-side previews per epoch; stop at the earliest one that nails it• Deliver to ComfyUI, and delivered files are named name_trigger.safetensors so the trigger is never a mystery• Duplicate images get caught — same-name doubles are flagged in the gallery with one-click delete• Feedback round 2: accurate speed/ETA dials, readable UI text (incl. Korean Windows), previews every epoch by default, config check at launch, side-by-side installs stay separate, real version stampEvery item here came from user feedback — keep it coming.Full detail: github.com/SillySilk/AnimaForge

AnimaForge Anima Lora Trainer

&lt;!-- CIVITAI VERSION UPDATE — AnimaForge v1.8 (user feedback round 2) HOW TO USE: Civitai → your AnimaForge model → Add Version (or edit the current one). • Version name: v1.8 — Community Feedback Round 2 • File: marketing/Zip file 2026-07-02/<a target="_blank" rel="ugc" href="http://AnimaForge.zip">AnimaForge.zip</a> (rebuilt 2026-07-02 from public main @ 94ef27e — replace the old zip) • Paste the "Version changelog" block below into the version description. • Optionally append the same block to the model description under a "Changelog" heading so it's visible without expanding versions.--&gt;# Version changelog — v1.8 (2026-07-02)Community feedback round 2 — every item shipped. GitHub always has the latest:<a target="_blank" rel="ugc" href="https://github.com/SillySilk/AnimaForge">https://github.com/SillySilk/AnimaForge</a>### 📊 Accurate speed / ETA dials (real bug, squashed)The progress readout was being polluted by other progress bars in the training log —every time a preview set rendered, the sampler's numbers briefly hijacked the stepcount and the Speed/ETA dials. The dials now only listen to the actual training bar,and the Speed dial auto-scales for fast cards (RTX 5090s pegged the old one).### 📁 Side-by-side installs stay separateTesting a new copy of AnimaForge no longer inherits the old copy's sd-scripts andoutput paths. Each install now uses its own bundled sd-scripts and output folder;deliberately external paths (your own kohya clone, an output drive) are still honored.### 🖱️ Click the tiles on HomeThe TARGET STEPS and OPTIMIZER tiles are buttons now — click them to open the StepCalculator / Train Presets directly. The Home summary also refreshes the moment youclose a Presets panel (it used to lag until you switched tabs).### 👀 Config check right before launchThe Start Training confirmation now includes the preview schedule and carries theexact TOML files under Show Details… — inspect the full recipe at the last moment.### 🖼️ Previews every epochSample previews default to every epoch again (was every 10, which hid progress forwhole runs). Watch the previews, stop at the earliest-best — adjustable in Presets.### 🎨 Readable UI text (incl. Korean Windows)Buttons, inputs, paths, labels, logs, and status now use a clean system font withproper CJK fallback; the forge character stays in the headlines and nav. Plus aspacing pass on the tightest panels, and a wider two-column Train Presets layout.### 📤 Deliver to ComfyUISet your ComfyUI loras folder in Setup and one click copies the trained LoRA there.Delivered files (Forge and ComfyUI) are also named *<code>{name}_{trigger}.safetensors</code>**now, so anyone you share the file with can read the trigger word right off it.### 🔢 Version stamp fixedThe sidebar now shows the real version (v1.8.0) instead of a stale v1.0.0.---*100% free &amp; open source (MIT). Keep the feedback coming — this release exists becauseyou sent it.*

Files with SHA256: 188c395fe41cebe3498e86d21e19fdc1dbaa443a63b9fea0f0b7e18a4135d00d