Skip to content

Vibe your highlights

Our project's goal is to generate customized sports highlights from video of events.

Code Video Cerebral Valley

Core trick

Voxtral timestamps the transcript, an LLM maps the user's request to matching time ranges, and ffmpeg trims and stitches them.

Tags

voicecreative/music

Description

Our project's goal is to generate customized sports highlights from video of events. For example, our user would love to get highlights of its favorite player: Luka Doncic, that's why he needs to complete information about the game, upload the video of the game and then it generates a video that combines all actions of Doncic during the game. Our project could be in the future extended to multiple sports (football, rugby, american football, cricket).
thomasinovic/gemini3-hackathon★ 11 forksPythonCreated Mar 14, 2026 · last push Mar 14, 2026

README excerpt

# NBA Custom Highlights Generator

## Overview
This project takes a full NBA match video and a natural language user prompt (e.g., "Show me all of Steph Curry's three-pointers" or "Give me the best defensive plays of the 4th quarter") and automatically generates a customized highlight reel.

## Core Workflow
1. **Transcription**: Extract audio from the full match and generate timestamped transcripts using the Mistral API (Voxtral).
2. **Analysis**: Feed the user's prompt and the timestamped transcript into an LLM to identify relevant events and their exact timestamps.
3. **Video Processing**: Use standard video editing libraries (e.g., `ffmpeg` or `moviepy`) to trim and stitch the identified segments into a single cohesive highlight video.

## Project Structure
- `config.py`: Centralized configuration (API keys, default paths, model names).
- `Design.md`: Detailed architectural design and feature documentation.
- `transcribe_voxtral.py`: Module for extracting audio and fetching transcripts via Mistral.
- `analyze_prompt.py`: Module for cross-referencing user prompts with transcripts using an LLM.
- `video_editor.py`: Module for cutting and stitching video segments.

## Setup
1. Install dependencies: `pip install -r requirements.txt` (ffmpeg must also be installed on your system).
2. Add your API keys to `config.py`.
3. Run the scripts directly (most include a `dry_run` test mode).

## Download 2024 Finals Game 5 Play-by-Play
This downloads the play-by-play JSON for **Dallas M

Team