VIBEMOTION / LEARN
How VibeMotion uses AI — and where Claude fits in
How VibeMotion uses language models like Claude to plan motion video copy, what AI controls vs. what you control, and how to configure Claude or other models.
The vibe motion workflow
Vibe motion describes a workflow where you give a plain-language description, and an AI language model turns it into a structured scene plan — headlines, copy, pacing — that a rendering engine brings to life. VibeMotion is a specific product built on this idea. You describe what you want; the model plans the copy; you stay in control of every scene.
How VibeMotion uses Claude
When AI planning is configured, VibeMotion sends your project description to a language model. The default path uses Claude — a language model made by Anthropic — via either the direct Anthropic API or OpenRouter. Claude reads your brief and returns a scene-by-scene plan: one headline, short supporting copy, and a motion speed for each scene. Nothing is invented: the model is instructed to use only the facts you supply and never add dates, statistics, or claims.
What Claude controls, and what you control
Claude writes a first draft of the scene copy within a fixed scene structure. You control: the words (you can edit every line manually for free), brand color, uploaded images, timing, and the final preview. The AI cannot change scene count, replace images, or alter the renderer. You can undo any AI change and restore a previous version.
Configuring Claude or another model
Set ANTHROPIC_API_KEY in your environment to use Claude directly — claude-haiku-4-5 for fast planning or claude-sonnet-4-5 for richer copy. Alternatively, set OPENROUTER_API_KEY with OPENROUTER_MODEL=anthropic/claude-haiku-4-5 to route through OpenRouter. If neither key is set, VibeMotion falls back to template mode: the same scene structure without AI copy.
Why use a language model at all
A first draft of copy is often the hardest part. The model turns a rough brief into readable, structured scene text in seconds, so you spend your time refining rather than starting from nothing. The result is always editable; the AI is a starting point, not the final word.
What this release cannot do
Silent 10–20 second designed motion scenes only. No photorealistic footage, no voiceovers, no music, no arbitrary code generation, no long-form video. Claude writes copy within a fixed scene template — it does not write code, generate images, or control the renderer directly.
How this compares to other tools
Runway and Kling generate photorealistic footage from text. CapCut and Canva offer template editors with manual controls. VibeMotion sits in between: an AI-planned, editable motion story from plain language, without needing to learn editing software or write code. The AI role is specifically copy planning, not video generation.