OrkasVideoStudio

OrkasVideoStudio lets a coding agent compose, edit, and generate video. You describe the video in plain language, and the agent reads the material, writes a timeline, and produces the file. Any agent that can run a shell or speak MCP can use it, including Claude Code, Codex, and Cursor.

The agent does the reasoning, and the project supplies three things: skills that encode what makes a good video and which production line to take, deterministic capabilities for rendering, editing, transcription, and generation (thin wrappers over HyperFrames, ffmpeg, and whisper.cpp), and an editable plan file. The plan is a readable, diffable intermediate representation of the video, so you or the agent can change one segment, caption, or narration line and only that piece re-renders.

Work splits into four production lines. Compose turns a script into designed HTML motion graphics and then MP4, for explainers, kinetic typography, data visualization, and title cards. Edit cuts, joins, trims silence, removes filler words, mixes, burns in subtitles, dubs, and localizes footage you supply, and selects highlights from long recordings. Generate produces AI footage, images, and narration through your own provider keys. Auto combines the lines into one cross-modal plan, and a deterministic delivery guard checks the finished cut (duration, canvas, audio, loudness, voiced spans) before it ships.

Compose and edit need no paid keys. The generation skills treat billable calls as a budget: they aim to reach most of the result without generation, cap a run's generated shots and recurring characters, require the plan's cost estimate to match its generated segments exactly, and re-bill only the segment that changed on a revision. The project began as the built-in video agent inside Orkas and was extracted into an agent-agnostic toolkit.