Smart Extractor
Understanding a video and preparing useful references from it.
Read as Markdown ↗Use it for: Understanding a video and preparing useful references from it.
What it can do: Write a breakdown guided by your question, suggest up to 12 timestamped keyframes, preview/select frames, prepare them as image assets, and optionally prepare the source audio's first 15 seconds.
Inputs: One ready video asset or finished Video node. Analysis currently accepts videos up to two minutes and 25 MB.
Outputs: Connectable written analysis. Selected frames can accompany the analysis into downstream Video generation. To use frames in Image generation, connect the prepared frame asset cards as image references. Optional extracted audio supports compatible Video audio-reference models.
Try it: Connect a short harbor clip. Ask “Describe the character's action, camera movement and scene changes. Suggest useful keyframes.” Choose Analyze video, then inspect the written breakdown and timestamps.
Remember: Previewing a suggested frame does not upload it. Explicit preparation saves it; downstream generation can prepare remaining selected frames. Changing the source video requires fresh analysis/references. Extracted references guide a recreation, rather than copying an entire video.
Watch the setup
These short demos show setup or existing results. Generation and analysis are started separately after you review the inputs and displayed price or allowance.