We will be undergoing planned maintenance on January 16th, 2026 at 1:00pm UTC. Please make sure to save your work.

RaceGram

Turn raw race telemetry into viral social cards. RaceGram detects race highlights (fastest laps, overtakes, close battles) and auto-generates Instagram-style story/post cards ready to share.

Comment

Inspiration

Social platforms thrive on digestible highlights. Motorsport has incredible micro-moments (comebacks, epic overtakes, perfect laps) that are perfect for short-form shareable content — but harvesting them from telemetry is manual and slow. RaceGram automates detection and generation of attractive share cards so teams, drivers and fans can instantly publish highlights.

What it does

  • Ingests race telemetry, lap times, sector data and results CSVs.
  • Runs a set of highlight-detection algorithms (10+ algorithms) to identify moments such as fastest lap, best overtake, close battles, position gains, comebacks, perfect laps, etc.
  • Generates finished image cards using Pillow-style rendering templates with captions, hashtags and QR codes ready for Instagram/Twitter/Facebook export.
  • Provides a React front-end gallery and a card customizer for tweaks before export.

How I built it

  • Data processor parses incoming CSV feeds into a normalized structure.
  • Highlight detector implements multiple heuristics/algorithms to score candidate highlights.
  • Story generator formats a short headline/narrative for each highlight.
  • Card renderer uses templates (JSON for layout) and Pillow-like drawing to produce final image assets.
  • Frontend built with React + Vite; Tailwind (index.css) for styles and an export manager for multi-format outputs.

Challenges I ran into

  • Defining “highlight.” What feels like a highlight depends on context (race class, number of cars, track). We needed multiple heuristics with tunable thresholds.
  • Aesthetics + readability. Automatically fitting text and telemetry overlays on a card requires flexible templates and fallbacks.
  • Edge cases in telemetry. Missing telemetry or inconsistent timestamps break simplistic detectors — robust parsing is a must.

Accomplishments that I'm proud of

  • 10+ highlight detection algorithms implemented and packaged.
  • Configurable card templates (JSON) that let the same engine produce many visual styles.
  • Tests for the core pipeline (data processor, highlight detector, card renderer).

What I learned

  • Combining simple heuristics + layout templates is fast to iterate on and works well for shareable content.
  • Providing a small UI for manual edits (caption tweak, crop) increases adoption because automated cards rarely need zero tweaks.

What's next for RaceGram

  • Add short video/gif export (animated telemetry overlays) and deeper NLP-based captions.
  • Add a simple scheduled-publish integration (post to Instagram/Twitter APIs) and templates branded for teams.

Built With

Share this project:

Updates