Printable production templates that multimodal AI can read.
kami (紙 paper / 神 divine) + tell (teach the agent)
kamitell provides A4-printable template sheets for creative production workflows:
- Screenplays / scripts
- Storyboards
- Character standing poses (full body)
- Face close-ups
- Scene layouts
Each template has labeled regions and guide lines. You print them, draw or write by hand, then scan. When a multimodal AI reads the scan, the pre-printed labels tell it exactly what each region contains — no explanation needed.
Some things are better drawn by hand. Spatial thinking, quick iteration, the texture of a pencil on paper — digital tools can't fully replace that. But hand-drawn work has always been hard to feed into digital pipelines.
kamitell bridges that gap. The templates act as a background layer (like a 下敷き) that gives structure to freeform hand work, while making the output immediately parseable by AI agents.
Print template → Draw/write by hand → Scan → AI reads labeled regions
This is not Figma. This is not Canva. This is for people who think with a pen in hand.
storyboard-portrait.pdf— Storyboard template (portrait A4)
More templates coming.
Bulk-converts scanned images to searchable Markdown using Tesseract. Designed for the "scan → text" step of the kamitell workflow.
# Single folder
./tools/ocr_batch.sh ./scans ./output
# Recursive (one output file per subfolder)
./tools/ocr_batch.sh ./scans ./output --recursive
# English only
./tools/ocr_batch.sh ./scans ./output --lang engOutput format: Numbered entries (### 001, ### 002, ...) separated by ---. Filename index table at the end. Clean to read, traceable by number.
CJK inter-character spaces are automatically removed (大 阪 府 → 大阪府).
Prerequisites: tesseract, tesseract-data-jpn, tesseract-data-eng, Python 3
Standalone post-processor for Tesseract's Japanese output. Removes the half-width spaces that Tesseract inserts between every CJK character.
python3 tools/fix_cjk_spaces.py ./output/
python3 tools/fix_cjk_spaces.py ./output/file.mdMIT