SecretSky

AI Video Agent for Short Films
Worldwide--:-- UTC
Scroll

Collections

Browse by collection

Curated sets of stills, grouped by the world they belong to and the way they're treated.

One agent.The whole film.

Start with an idea, a story, or a visual reference. Secret Sky develops it into scenes, shots, and a production plan—while keeping the whole film connected as it changes.

Develop the story · Break down shots · Direct in 3D · Rehearse · Move into production

Direct in a stage you can see.

Tell the agent: “Have her walk to the window, turn back, and hold as the camera slowly pushes in.” It edits the visible 3D character, pose, motion, path, camera, and timeline—then rehearses the result.

  • Character
  • Pose
  • Motion
  • Path
  • Camera
  • Timeline
  • Rehearse

Start from the reference you already have.

Build from natural language, reconstruct an image as an editable 3D blocking reference, or extract depth, pose, and motion structure from a reference clip.

Bring Claude or Codex into the director’s chair.

Through Stage MCP, Claude or Codex can read the scene, direct character motion and cameras, edit the timeline, switch scene states, and rehearse the result. Every action remains visible, editable, and traceable in Canvas.

Every decision stays with the film.

The Film Agent keeps shots, assets, dependencies, approvals, costs, and results in one project. When a key decision changes, affected work stays visible—and generation or spending remains under human approval.

FAQ

A clearer point of view.

Secret Sky is a studio that takes AI generations to finished, production-grade work — across image, motion, character, and storyboard. Selected methods are published as Skills.

Generating an image or a video is easy now. Getting one good enough to ship — cinema-grade, consistent, finished — is not. That second part is what Secret Sky does.

Secret Sky releases selected, tested creative methods as Skills for image direction, character design, storyboarding, image-to-video, and visual structure. Only Skills explicitly labeled open source are open source; every item states its inputs, outputs, use cases, status, and access model.

Models produce output; they don't know what's right for an idea. Composition, light, rhythm, consistency — that judgment is the method, and the difference between a generation and finished work.

You can begin with a sentence, a feeling, a story, an image, or a reference film. Different Skills turn those inputs into images, characters, storyboards, motion direction, or shot structures that can be developed further.

Depending on the task: Kling, Veo, Seedance, GPT Image, Gemini Image. The model is creative material — direction and quality come from method, not the model name.

Begin

Start with an idea.

A sentence, a feeling, or a reference can be enough. Give it image, structure, and texture—and let it grow into a work.

Or write directly · leo414x@gmail.com