Inspiration

Every content creator knows the struggle: you spend hours filming and editing, only to hit a wall of manual chores before hitting publish. Captioning, writing SEO metadata, designing A/B thumbnails, and analyzing past retention metrics take hours of spreadsheet work and guessing. I wanted to build a tool that completely automates this pipeline from a single transcript, completely offline.

What it does

Creator Toolkit is a zero-API, local Python pipeline and interactive dashboard that solves three major pre-upload bottlenecks and one major post-upload bottleneck:

  • Automated Captions: Generates precise .srt files from plain text scripts.
  • SEO Engine: Extracts keywords locally using YAKE, then auto-generates titles, tags, hashtags, and a formatted YouTube description with chapters.
  • Thumbnail Generator: Renders A/B testable 720p thumbnails with dynamic color palettes and bold text overlays using Pillow.
  • Streamlit Analytics Dashboard: A visual triage tool that ingests raw YouTube Studio CSV exports to instantly flag "30-second cliffs," calculate optimal posting schedules based on CTR, and score audience sentiment.

How I built it

I built the core engine in pure Python, prioritizing an offline-first, zero-API approach so creators don't have to worry about rate limits or network calls.

  • I utilized YAKE for unsupervised statistical keyphrase extraction.
  • I used Pillow for programmatic image and text rendering.
  • I built the interactive front-end using Streamlit, leveraging Pandas and NumPy for rapid CSV data processing and TextBlob for NLP sentiment analysis on audience comments.

Challenges I ran into

Building a keyword extractor that didn't rely on massive external LLMs was tough; I initially tried a hand-rolled RAKE variant but pivoted to YAKE to handle punctuation-sparse speech transcripts better. Additionally, standardizing raw YouTube Studio CSV exports (which change formatting based on locale and user settings) required building robust fallback logic in my Streamlit app.

What's next for Creator Toolkit

I plan to integrate local, offline ASR (like faster-whisper) so the tool can generate its own transcripts directly from raw video or audio files, removing the need for the user to provide a text script entirely!

Built With

Share this project:

Updates

Submission history