Character LoRA
Orchestrates the end-to-end pipeline to train a consistent-identity character LoRA on a local GPU
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 17, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · d2525e2
Task: plan a character LoRA for an original anime character on Pony V6 (caption format, dim/alpha, and the top mistake to avoid). I could not run the pipeline it orchestrates: it needs a local GPU, the codex/gpt-image-gen image stack, and a kohya/ai-toolkit trainer, none available here, so the headline deliverable (a trained LoRA) is unmeasured. On the planning slice I could test, the skill's guidance is specific and correct where a from-memory baseline stays generic: trigger as a non-dictionary first token with keep_tokens=1, alpha never above dim, no quality/score tags in training captions, the LoRA locked to its base architecture, and picking the best checkpoint rather than the last. It installs cleanly and honestly marks unverified settings as 'proposed', but without the GPU pipeline this is advisory only, hence setup rather than pass.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 6/10
- Docs & honesty 4/5
What Character LoRA does
Drives the multi-stage pipeline for building a consistent-identity LoRA of an original character: define the identity, generate a face/body-consistent multi-angle dataset (delegating image generation to gpt-image-gen), caption it per base, do base-specific homework, and train on a local GPU. Gates every expensive or irreversible step on user approval and refuses to improvise training settings from memory.
How to install Character LoRA
git clone https://github.com/KerberosClaw/kc_ai_skills
mkdir -p ~/.claude/skills
cd kc_ai_skills && cp -r character-lora ~/.claude/skills/character-lora
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Character LoRA
-
/character-loraOrchestrates the end-to-end pipeline to train a consistent-identity character LoRA on a local GPU
It also activates on plain-language prompts like these:
-
Plan a character LoRA training pipeline for our character on Pony V6 -
What caption format and alpha or dim should I use for a character LoRA -
Help me build a multi-angle dataset for training a new character's LoRA
Frequently asked questions
- Is the Character LoRA skill free?
- Yes. The skill itself is free from KerberosClaw/kc_ai_skills. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Character LoRA work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 17, 2026. Verdict: Works with setup. Task: plan a character LoRA for an original anime character on Pony V6 (caption format, dim/alpha, and the top mistake to avoid). I could not run the pipeline it orchestrates: it needs a local GPU, the codex/gpt-image-gen image stack, and a kohya/ai-toolkit trainer, none available here, so the headline deliverable (a trained LoRA) is unmeasured. On the planning slice I could test, the skill's guidance is specific and correct where a from-memory baseline stays generic: trigger as a non-dictionary first token with keep_tokens=1, alpha never above dim, no quality/score tags in training captions, the LoRA locked to its base architecture, and picking the best checkpoint rather than the last. It installs cleanly and honestly marks unverified settings as 'proposed', but without the GPU pipeline this is advisory only, hence setup rather than pass.
- What is the Character LoRA SkillProof Score?
- 8.0/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 6/10, docs & honesty 4/5.
- How do I install Character LoRA?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Character LoRA with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.