Let your AI fill the form, and keep the render button

Video has no AI of its own. It has a door for yours. Once Video is added to your Geenius OS workspace and your assistant is connected to Geenius OS, the assistant has a tool called geenius_video with five areas: template, video, collection, asset and snippet.

A good order of work#

1. Read the schema. Ask the assistant to read the template, which returns the schema and a valid sample. Starting from the sample, it only has to change values.
2. Create the draft. The first write validates the whole spec. If something is wrong, the message names the scene and the field and says what to write.
3. Edit scene by scene. Each edit is atomic. The assistant can add a scene, update one, move it or copy scenes from another video, without rewriting the rest.
4. Lint. It asks for the quality warnings and fixes them: too many words for the time, text outside the safe area, a scene with no motion.
5. Hand over the link. The draft waits until a person opens it in Video and presses Render. The assistant reads back the link and gives it to you.

What it cannot do#

It cannot render, because rendering runs in a browser on a person's machine. It cannot change the file of an imported film. It cannot make footage, because Video has none to make.

Tips for the prompt#

• Name the template and the size. "Release notes, 1080p" beats "a video about the release".
• Give it the real content and ask it not to invent features.
• Ask for lint and the link at the end.
• For a long video, ask for an outline first, then chapter by chapter.

A sample ask#

Use Geenius OS and my Video account. Make a quote video from this line: "Ship the small thing today." Use my quote template at square size, check it with lint and give me the link.

Why this split#

The assistant is good at words and structure. The preview is where you check taste. The render button is where the file becomes real. Keeping those three apart keeps you in charge of what leaves the building.