Literate AI

Repository documentation

Credits and asset sources

Every asset in Settle the Argument. Before CI. is listed here with its source and license. production.json in each authoring package repeats the exact fonts, tool versions and scene-script hashes used for that render.

Original to this repository

Asset Source Notes
LitAI and Sam portraits ../assets Fictional characters; no real likeness
3D scenes scenes/blender_scenes.py Generic shapes, no logos; labeled ILLUSTRATION in the film
Cards, captions, close-ups scenes/cards.py Composed from the portraits and text
Sound bed and UI cues scenes/sound.py Synthesized by ffmpeg formulas; no recordings or samples
Terminal and file pages collect.py evidence Excerpts of recorded runs and repository files

Third-party inputs

Asset Version License Use
Inter 4.1 (Inter-4.1.zip sha256 9883fdd4…b11e) SIL OFL 1.1 Captions, labels; Inter Display Bold for titles
JetBrains Mono 2.304 (JetBrainsMono-2.304.zip sha256 6f6376c6…7bbf) SIL OFL 1.1 Terminal pages
Qwen3-TTS CustomVoice, MLX 4-bit conversion snapshot f35faf19b0cc Apache-2.0 (model card) Local narration with built-in presets Serena and Ryan; no voice cloning, no cloud speech
TinyXML2 8224e427 zlib Filenames and harness evidence only; implementation not shown
Cytoscape.js (inside recorded litai render html output) 3.34.3 MIT Appears in screenshots of the rendered graph page

Fonts are downloaded into the private build directory, not committed; the film embeds rendered pixels, which the OFL permits. Model weights are never redistributed: only generated narration ships.

Tools (not redistributed)

Blender 5.2.2 LTS (3D rendering), ImageMagick 7.1.2, ffmpeg 9.0.2, mlx-audio 0.5.7, Playwright with the locally installed Chrome (screenshots of local HTML files). Speech-recognition checks (listen.py) use a local Whisper large-v3 turbo snapshot; its transcripts are QA notes, never part of the film.

Read the Markdown sourceConference homePublished 1.3.0 · Patch 1.3.1 in progress · Development 1.4