Some new features on RedBoy – Granular AI Video Editor, 5-Min V2V & Auto-Foley

When we first launched, if the AI generated a weird glitch in shot 12 of your 2-minute film, your only option was a basic "reprocess scene" button. You'd re-roll the prompt with new user direction and cross your fingers, hoping it wouldn't ruin the continuity.

Creators told us loud and clear: stop giving us black boxes. Give us control.

So we tore down our backend and built a real, granular video editor designed specifically for AI workflows. Here is what we just shipped:

The Granular Video Editor We built an in-app timeline editor that treats every generated scene as a modular node. The biggest update is in-scene entity editing—you can swap or edit specific characters and objects within a single scene without breaking the overall visual composition. You can also trim, add, or delete individual clips post-generation, and fine-tune Foley, swap multi-language TTS voices, and edit subtitles right on the timeline.

5-Minute Video-to-Video (V2V) You can now transform real videos up to 5 minutes long. To prevent timeouts and memory crashes, our engine automatically slices videos into editable chunks, maintains temporal consistency across cuts, and stitches them back together. You can even inject your own generated AI characters into real-world footage.

Cost-Effective Auto-Foley & Music Mode Top-tier video models with native audio are expensive. To keep generations fast and affordable, we pair silent models with a custom audio engine. It automatically detects visual actions to generate synced Foley and sound effects, or syncs visual cuts to custom music tracks up to 2 minutes long.

Smart Topic Feeds Instead of one chaotic feed, you can now discover community-created short films filtered by specific interests like History, Science, Anime, and Family.

WIP: MCP Server We are also currently building a MCP server. Soon, you'll be able to hook RedBoy's backend directly into your favorite LLMs to build storyboards, trigger animations, and make editor tweaks through natural conversation—similar to OpenArt's flow.

Building the orchestration behind granular entity swapping and long-form V2V chunking was a huge technical journey, but our goal is to give creators the precision they actually need to tell real stories.

I would love to get your thoughts. If you are a creator, what feature is currently missing from your AI video workflow? What are you most looking for in a tool like this?

Drop your opinions and feature wishlists in the comments, I will be hanging out here all day taking notes and answering questions!