Seedance Analytics

Analytics pipeline for Seedance 2 video sequence generation — prompt editing behaviour, per-user section changes, and Word report generation (2026).

Scripts

Script	Purpose
seedance_sequence_analytics.py	Core analytics: fetches all Seedance video generation jobs from MongoDB, correlates AI-generated prompts with the prompts actually sent to the model, computes edit rates/similarity ratios, and writes reports/seedance_analytics.json
dump_seedance_records.py	Dumps all paired Seedance jobs to a navigable HTML page and Markdown file (reports/seedance_records.html, seedance_records.md) for per-job inspection
plot_seedance_user_sections.py	Reads MongoDB section-edit events, classifies actions per user (constraints / lighting / artstyle changes), builds a D3 heatmap HTML and a PNG (reports/seedance_per_user_sections.html, .png, .json)
build_report_docx.py	Assembles the final Word report (reports/report.docx) from the JSON + PNG outputs of the above scripts

Setup

pip install "pymongo==4.17.0" "python-docx==1.2.0" "matplotlib==3.11.0" "numpy==2.5.1"

The scripts read the MongoDB connection string from a .env file (or MONGODB_URI env var) at the project root.

Usage

Run in order — each script writes outputs consumed by the next:

# 1. Core analytics → reports/seedance_analytics.json
python seedance_sequence_analytics.py

# 2. Per-job HTML/Markdown dump → reports/seedance_records.*
python dump_seedance_records.py

# 3. Per-user section heatmap → reports/seedance_per_user_sections.*
python plot_seedance_user_sections.py

# 4. Word report → reports/report.docx
python build_report_docx.py

All scripts are read-only against MongoDB. Outputs land in a reports/ directory one level above the script location.

Engineering exclusion

seedance_sequence_analytics.py excludes jobs where the user or story owner has an @scenarix.ai email. The per-user section and report scripts apply the same exclusion via a shared EXCLUDED_REPORT_EMAILS constant.