How flows work: nodes, artifacts, and runs
The mental model behind every Hyperclip video. Nodes produce artifacts; runs walk the graph; the final composition is rendered once at the end.
A flow looks like a graph because it is one. Understanding three concepts — nodes, artifacts, and runs — is enough to confidently build any flow on Hyperclip.
Nodes are steps
Every box on the canvas is a node. A node has inputs, does one thing — generate a script, generate an image, add captions, mix in music — and emits an output. The arrows between nodes describe the order in which they run.
Artifacts pass between nodes
When a node finishes, it produces an artifact: a typed bundle carrying things like a script, an image, a video clip, caption timings, or an audio track. Downstream nodes pick out what they need and pass the rest along untouched.
This is why you can swap an image node for a different one without rewiring the captions and music — they're reading from the artifact, not from a specific upstream node.
A run walks the graph
When you hit Run, Hyperclip walks the graph from inputs to outputs, executing each node in order and accumulating the composition as it goes. Voiceover, captions, and music nodes layer onto the artifact without producing intermediate video files.
One render at the end
At the end of the run, Hyperclip assembles the layers — visuals, voice, captions, music — and renders the final video once. One render per run means one predictable cost, regardless of how many intermediate nodes the flow has.
Why this matters
It's the same model whether you run a flow from the canvas, schedule it for nightly batch runs, or call the public API. The graph you tested is the graph that ships. No second engine, no surprise diffs between what you previewed and what you get charged for.
