Scrambo

Build your own AI video team

Scrambo is a Python framework for building reliable video-agent workflows with Claude, Codex, or your preferred coding agent. You define how the team plans, edits, reviews, and improves; Scrambo handles the timeline, agent runtime, perceptual validation, and rendering underneath.

Private beta — invite only

Creative Hook for Social Reels

Animated grids, SFX, fast montage cuts, typography layout

Craft specialized video agents in minutes

Scrambo is not just one agent, it's a team of specialist video agents that you can command. Use chain, loop, or graph to coordinate them. Turn your ideas into production-ready workflows.

Asset generation agent

Describe asset ideas, give agent generative tools, control budget.

brief = source.generate_agent.create("Do: 1) Create a narrated 16:9 tour from the listing photos of this home. Turn the strongest stills into slow, low-motion 6-second push-ins (24mm interiors, 28mm exteriors) that preserve the exact architecture. 2) Write a warm, concise narration that walks a buyer from the front exterior through the interior to the deck and backyard.",tools=[img2video, voiceover],budget_usd=5.0,)
Planning agent

Direct the narrative spine of the edit — what to show, when, and for how long.

plan = timeline.storyboard_agent.plan(brief,"Plan a real-estate listing reel that matches the length of the voiceover. Compute the voiceover transcript so section timing follows the narration's phrase boundaries. Cut on narration phrase boundaries with clean hard cuts, no dialogue other than the voiceover. If background_music is used, keep it as a very low bed under the narration.",tools=[transcribe],)
Editing agents

Assembles the edits (cuts, pacing, captions, effects) into a finished reel.

timeline.author_agent.edit(plan) timeline.captions_agent.edit("Add transcript-aligned captions for the voiceover narration, styled as a premium real-estate reel. Reveal one word at a time in sync with speech, and fade each word in as it appears...",)
Review and refine in the editor

Make adjustment in Scrambo cloud editor.

Quality guardrails

Agents keep iterating on the edit until it meets your quality standards. Build your own policies, or use Scrambo's perceptual and structural quality checks.

timeline.ensure("edit_quality.pacing, edit_quality.structure",repair_with=timeline.author_agent,require_passed=True,)timeline.ensure("edit_quality.caption, edit_quality.typography",repair_with=timeline.captions_agent,require_passed=True,)
Export

Bring your AI video ideas, leave the infrastructure to us

Most AI video tools are a closed app that renders one clip from a prompt. Scrambo lets you create these tools yourself by chatting with your existing coding agent.

Own your video workflow

Other tools lock you into one app and one way of making video. Scrambo hands you editing and agent primitives in Python, so you turn your own production process into custom video workers — and own the workflow instead of assembling the underlying video stack.

Direct specialists, not prompt-and-pray

A single prompt to a single model is a single guess. Scrambo lets you delegate planning, analysis, editing, and review to focused agents with clear responsibilities and handoffs, composed into loops, graphs, or custom workflows that match how your production actually runs.

Built for reliability

One-shot generators ship whatever comes out. Scrambo agents watch the edit, evaluate whether it actually works, and revise it when it doesn't — with perceptual and structural checks that turn fragile prompts into repeatable production workflows.

Reusable and extensible

A generated clip is a dead end; a Scrambo workflow is a Python program you can inspect, version, test, and reuse. Bring your own models, media tools, APIs, and business logic, then turn one workflow into your own video vertical.