How dleungdl/learntok works
LearnTok AI turns learning material or a script JSON into a vertical two-character educational video. The primary workflow is optional LLM script generation and quality checks, followed by Edge-TTS, optional RVC voice conversion, loudness calibration, and ffmpeg composition. Other documented branches include series planning and approval, RAG knowledge-base build/retrieval, subtitle ingestion, term migration, and environment checks. Edges follow the README workflow or explicit calls in sampled source; unsampled subsystems are represented only where their paths are in the tree and their capabilities are documented.
Main entry points: compose.py Video composition · script_gen.py Script generation · run_pipeline.py Pipeline runner · tts_edge.py Edge TTS
Components 25
Script authoring
- Script generation
src/learntok/tools/script_gen.py - Script quality gate
src/learntok/tools/validate_script.py - Deterministic script fix
src/learntok/tools/script_fix.py - SRT ingestion
src/learntok/tools/ingest_srt.py - Term migration
src/learntok/tools/migrate_terms.py
Knowledge and series
- Series planning
src/learntok/tools/series_bible.py - RAG indexing
src/learntok/tools/rag_build.py - ChromaDB knowledge base
src/learntok/tools/rag_common.py - RAG retrieval
src/learntok/tools/rag_retrieve.py
Audio and video
- Pipeline runner
src/learntok/tools/run_pipeline.py - Edge TTS
src/learntok/tools/tts_edge.py - RVC conversion
src/learntok/tools/rvc_convert.py - Loudness calibration
src/learntok/tools/calibrate_audio.py - Video composition
src/learntok/compose.py - Rendered video
Workspace and safety
- Workspace configuration
src/learntok/config.py - Model integrity checks
src/learntok/tools/model_safety.py - Environment doctor
src/learntok/doctor.py - Characters and media assets
assets/manifest.json
Other components
- Creator
- Learning material
- LLM provider
- Edge TTS service
- ffmpeg
- Video output
Connections 32
- Creator → Learning material: provides
README.md - Creator → Script generation: requests generation
README.md - Learning material → Script generation: supplies source
README.md - Script generation → LLM provider: prompts
README.md - Script generation → Script quality gate: validates output
README.md - Script quality gate → Deterministic script fix: fixes issues
src/learntok/tools/script_gen.py - Deterministic script fix → Script quality gate: rechecks
src/learntok/tools/script_fix.py - Creator → Pipeline runner: starts render
README.md - Pipeline runner → Edge TTS: runs TTS
src/learntok/tools/run_pipeline.py - Pipeline runner → RVC conversion: runs optionally
src/learntok/tools/run_pipeline.py - Pipeline runner → Loudness calibration: runs calibration
src/learntok/tools/run_pipeline.py - Pipeline runner → Video composition: runs composition
src/learntok/tools/run_pipeline.py - Edge TTS → Edge TTS service: synthesizes speech
README.md - Edge TTS → Video composition: writes timed audio
src/learntok/tools/tts_edge.py - RVC conversion → Model integrity checks: verifies models
src/learntok/tools/rvc_convert.py - Loudness calibration → Workspace configuration: resolves workspace
src/learntok/tools/calibrate_audio.py - Video composition → Workspace configuration: resolves tools
src/learntok/compose.py - Video composition → Characters and media assets: reads media manifest
README.md - Video composition → ffmpeg: renders with
README.md - Video composition → Video output: writes video
README.md - Video output → Rendered video: contains
README.md - Creator → Series planning: plans learning path
README.md - Series planning → Script generation: generates episodes
src/learntok/tools/series_bible.py - Creator → RAG indexing: builds knowledge base
README.md - RAG indexing → ChromaDB knowledge base: upserts chunks
src/learntok/tools/rag_build.py - RAG retrieval → ChromaDB knowledge base: queries chunks
src/learntok/tools/rag_retrieve.py - Script generation → ChromaDB knowledge base: retrieves context
src/learntok/tools/script_gen.py - Creator → RAG retrieval: searches facts
README.md - Creator → SRT ingestion: imports subtitles
README.md - Creator → Term migration: migrates terms
README.md - Environment doctor → Workspace configuration: checks environment
src/learntok/doctor.py - Environment doctor → Model integrity checks: checks models
src/learntok/doctor.py