← All projects

Client work · Web

Wandera AI

AI video editor that turns raw footage into videos people want to watch.

Wandera’s editor: a video preview above a layered timeline with captions, and an “Edit with AI” chat that restyles the footage on request.
A Wandera project where edits are requested in chat, with saved versions to compare and the AI explaining its choices.
Wandera’s home screen: “What are we making today?”, a prompt box to describe an edit and attach a clip, and recent projects.
Wandera’s style gallery: ready-made looks such as Podcast Frame, Breaking News, Quote Card, Tutorial Steps, and Countdown.
Wandera’s landing page: “Say the edit. See it happen.” beside the editor, with an AI chat panel and a layered timeline.

01 / 05Timeline editor, with edits made through chat

About the app

  • Describe an edit in plain words and watch it happen
  • Refine every detail in a real, layered timeline editor
  • Captions, pacing, B-roll, zooms, and music from one prompt

What I worked on

  • Built the Flutter app, including the layered timeline editor architecture
  • Remotion Lambda render pipeline
  • Gemini Omni integration via Vertex AI
  • Benchmarking frontier models on video-editing tasks (the Wandera RL benchmark)

Built with

  • Flutter
  • Remotion Lambda
  • Gemini Omni via Vertex

What I’ve been working on lately

Wandera RL benchmark

A benchmark of video-creation skill at the edge of model ability. It turns the Wandera editor into a reinforcement-learning environment: models work through typed editor tools, a verifier checks the finished project, and every run earns a reward from 0 to 1. Frontier models from OpenAI, Anthropic, Google, and xAI are scored on hard tasks, such as rebuilding real product-launch videos, with public traces and findings.

View the benchmark
The Wandera RL benchmark leaderboard: frontier models ranked by mean reward across seven hard video-editing tasks.