How aryanpatil4506/dropzero works
DROPZERO is a local-first video audience simulator. A creator uploads a video or script, starts analysis, and receives a transcript, timed segments, multimodal features, retention predictions, evidence-backed drop-off flags, and edit suggestions. The dashboard also supports edit-impact simulation, rendering edited copies, A/B comparisons, and validation against real viewing data. The map combines the README workflow with the sampled API, pipeline, detection, rendering, and frontend files; several unsampled subsystems are represented at their exact directories, so some internal links are intentionally omitted.
Main entry points: runner.py Analysis runner · projects.py Project API · client.ts API client · Dashboard.tsx Project dashboard
Components 21
Creator experience
- Project dashboard
frontend/src/pages/Dashboard.tsx - API client
frontend/src/api/client.ts
Analysis pipeline
- Analysis runner
backend/app/pipeline/runner.py - Media ingestion
backend/app/ingestion - Transcription
backend/app/transcription - Timed segmentation
backend/app/segmentation/windows.py - Text features
backend/app/features/text/extract.py - Audio-visual features
backend/app/features/av.py - Local model runtime
backend/app/features/text/embedder.py
Retention insights
- Retention prediction
backend/app/model/predict.py - Drop-off detection
backend/app/detection/flags.py - Title-promise check
backend/app/detection/promises.py - Evidence explanations
backend/app/explain/llm.py
Edits and validation
- Edit simulation
backend/app/simulate - Edited-video rendering
backend/app/render/service.py - Model validation
backend/app/validation
API and storage
- Project API
backend/app/api/projects.py - Job API
backend/app/api/jobs.py - Project and artifact store
backend/app/storage/db.py - Encrypted media store
backend/app/storage/media_store.py
Other components
- Creator
Connections 23
- Creator → Project dashboard: uses
README.md - Project dashboard → API client: requests data
frontend/src/pages/Dashboard.tsx - API client → Project API: calls project endpoints
README.md - API client → Job API: polls jobs
frontend/src/api/client.ts - Project API → Analysis runner: starts analysis
backend/app/api/projects.py - Analysis runner → Media ingestion: extracts media
backend/app/pipeline/runner.py - Analysis runner → Transcription: transcribes
backend/app/pipeline/runner.py - Analysis runner → Timed segmentation: builds segments
backend/app/pipeline/runner.py - Analysis runner → Text features: extracts text features
backend/app/pipeline/runner.py - Analysis runner → Audio-visual features: extracts AV features
backend/app/pipeline/runner.py - Text features → Local model runtime: uses embeddings
backend/app/features/text/extract.py - Analysis runner → Retention prediction: predicts retention
backend/app/pipeline/runner.py - Analysis runner → Title-promise check: checks title promise
backend/app/pipeline/runner.py - Analysis runner → Drop-off detection: detects drop-offs
backend/app/pipeline/runner.py - Drop-off detection → Evidence explanations: explains flags
backend/app/explain/llm.py - Project API → Edit simulation: simulates edits
README.md - Project API → Edited-video rendering: requests renders
backend/app/api/projects.py - Edited-video rendering → Encrypted media store: stores encrypted output
backend/app/render/service.py - Analysis runner → Project and artifact store: persists artifacts
backend/app/pipeline/runner.py - Project API → Project and artifact store: reads project data
backend/app/api/projects.py - Project API → Encrypted media store: serves uploaded media
- API client → Model validation: fetches validation
README.md - Retention prediction → Model validation: evaluated against viewing data
README.md