skill

fal-portrait-lora-skills

Cursor Agent Skills for preparing portrait LoRA datasets and training them with fal-ai/flux-lora-portrait-trainer

shahshrey/fal-portrait-lora-skillscommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/shahshrey/fal-portrait-lora-skills itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 21, 2026

fal Portrait LoRA Skills

Open-source Cursor Agent Skills covering the whole portrait LoRA loop: prepare a dataset, train it with fal-ai/flux-lora-portrait-trainer, then generate and evaluate images with fal-ai/flux-lora.

Skill	What it does
`lora-dataset-prep`	Face-centered 1024×1024 crops, contact-sheet visual QA, Florence-2 captions, flat ZIP packing
`fal-lora-training`	Validate ZIP → upload → submit → monitor → persist LoRA weights
`fal-lora-inference`	Generate from trained weights, sweep LoRA scale, download images, contact sheet for review

These skills teach an agent the workflow end-to-end, and ship the Python helpers the agent runs.

Install

Cursor (recommended)

Clone or download this repo.
Copy (or symlink) the skill folders into your Cursor skills directory:

# Personal (all projects)
mkdir -p ~/.cursor/skills
cp -R skills/* ~/.cursor/skills/

# Or project-local
mkdir -p .cursor/skills
cp -R skills/* .cursor/skills/

In Agent chat, invoke with /lora-dataset-prep, /fal-lora-training, or /fal-lora-inference — or just ask to prepare, train, or test a portrait LoRA, and the agent should pick them up from the skill descriptions.

One-liner helper

./scripts/install.sh            # installs into ~/.cursor/skills
./scripts/install.sh --project  # installs into ./.cursor/skills

Quick start

# 1) Dataset prep env + YuNet model
bash ~/.cursor/skills/lora-dataset-prep/scripts/setup_env.sh

# 2) Crop every source photo for review
.venv/bin/python ~/.cursor/skills/lora-dataset-prep/scripts/prepare_all_portraits.py \
  --src "/path/to/raw_photos" \
  --out "./prepared_dataset" \
  --model "./models/face_detection_yunet_2023mar.onnx" \
  --prefix "subject"

# 3) Visually QA contact sheets, manually recrop failures, then caption + zip
# (see skills/lora-dataset-prep/SKILL.md)

# 4) Train (paid fal job — requires explicit confirmation / --yes)
.venv/bin/python ~/.cursor/skills/fal-lora-training/scripts/train_portrait_lora.py \
  --archive "./prepared_dataset/dataset.zip" \
  --trigger-phrase "ohwx man" \
  --output-dir "./portrait_training_run" \
  --env ".env" \
  --yes

# 5) Test the LoRA across scales (paid fal job) and review the contact sheet
.venv/bin/python ~/.cursor/skills/fal-lora-inference/scripts/generate_lora_images.py \
  --lora "./portrait_training_run/subject_lora.safetensors" \
  --trigger-phrase "ohwx man" \
  --prompt "ohwx man, close-up portrait, soft window light, navy blazer" \
  --lora-scale 0.7 0.9 1.1 1.3 \
  --seed 12345 \
  --output-dir "./lora_generations" \
  --env ".env" \
  --yes

The trigger phrase must appear in every prompt — it is what carries the identity. A prompt without it renders a generic person even with the LoRA loaded.

Auth for captioning, training, and inference: set FAL_KEY or FAL_API_KEY in a local .env (never commit it).

Repository layout

skills/
  lora-dataset-prep/     # dataset prep skill + scripts
  fal-lora-training/     # fal training skill + scripts
  fal-lora-inference/    # generation + evaluation skill + scripts
scripts/
  install.sh             # copy skills into Cursor
  check.sh               # smoke checks (same script CI runs)
.github/                 # issue/PR templates, CI, CODEOWNERS

Requirements

Python 3.10+
Cursor (for skill invocation) or any agent that loads Agent Skills (SKILL.md)
Optional: fal account + API key for captioning and training

Contributing

Contributions are welcome — bug fixes, better crop heuristics, caption polish, docs, and new related skills.

See CONTRIBUTING.md for the workflow, coding standards, and review expectations.

Please read the Code of Conduct before participating.

Security

Do not open issues with API keys, private photos, or training URLs you consider sensitive. Report security issues privately per SECURITY.md.

License

fal.ai, FLUX, and Cursor are trademarks of their respective owners. This project is not affiliated with or endorsed by fal, Black Forest Labs, or Anysphere.