Dataset Tools

mere-dataset-tools prepares local image datasets for LoRA workflows with captions, optional OCR sidecars, trigger tokens, focus guidance, and a contact sheet.

Install and caption

mere.run plugin install mere-dataset-tools
mere-dataset-tools doctor
mere-dataset-tools caption \
  --input ./dataset \
  --output-dir ./caption-out \
  --trigger-token STYLE \
  --focus "material and silhouette" \
  --contact-sheet

Add --ocr when visible text should be extracted into sidecars. Use --prompt to supply caption guidance and repeat --focus for multiple priorities.

Plan first

mere-dataset-tools plan \
  --input ./dataset \
  --output-dir ./caption-out \
  --trigger-token STYLE \
  --run-id dataset-001
mere-dataset-tools run ./caption-out/run.json

Next step

Inspect the captions and contact sheet before training. A prepared paired dataset can feed the Klein LoRA recipe through RunPod Runner.