Video Timeline (Text)

Modern video models are getting good at following a sequence - but only if you tell them the sequence. That's what this node is for: it turns "5 shots over 5 seconds" into a ready-to-paste timeline like Shot 1 at 0.000s. Shot 2 at 1.250s. and so on, so the model knows what's supposed to happen when. It's the text-only sibling of the pack's image-driven timeline nodes: no images in, just structured guidance out, built from a duration you type or a video you connect.

This is the node for the family of video prompts that take timestamped shot guidance - the MiniMax H3 workflows in this same pack being the obvious target, where "Picture timestamps map one-to-one to already-selected images" and the text carries the timeline. If your workflow already has the images, you don't need this node. If you're prompting a video model purely from text, this is your shot list.

How it works

You set a duration in seconds, and segment_count (default 5) decides how many timestamped entries to lay across it. The outputs are:

• timestamps text - every timestamp in one line.
• timeline text - each entry formatted via timeline_text_structure, default Shot <<shot>> at <<timestamp>>. You can swap in <<time>> or <<timestamp>> placeholders and any prose you like.
• structured timeline text - one whole-paragraph template, default Target video duration is <<duration>> seconds divided into <<segments>> segments. <<shot>> at <<timestamp>>., where <<duration>> and <<segments>> are scalars and the shot/timestamp section repeats for every entry.
• video runtime - the length in seconds as a number, handy if you're driving other nodes off it.

The clever bit is focus_areas (0–3) plus focus_one, focus_two, focus_three. With focus_areas at 0 (default), entries space evenly. Set it to 1 and all entries cluster toward one end of the timeline - the focus values run 0 (early) to 1 (late), 0.5 balanced. Two or three parts lets you cluster different groups at different points: shots bunched in the opening, a gap, shots bunched in the finale. It's a compact way to express pacing without hand-authoring every timestamp.

Two things worth knowing

• Connect the optional video input and its native duration overrides your duration widget - the widget is only for manual use. Useful when you're building a timeline to match footage you already have.
• timestamp_format (default 00.000s) is reused verbatim in every output, so change it once and all four outputs follow.

The trap to avoid

This is a text templating node, so the failure mode isn't a crash - it's feeding garbage structure to a model that takes it literally. If your video model ignores the timeline entirely, check that the template placeholders are spelled right (<<timestamp>>, not {timestamp}), and remember focus values only matter for the parts they belong to: focus_two does nothing unless focus_areas is at least 2. It won't generate a great video by itself - nothing will - but it's the fastest way to hand a model a pacing plan it can actually read.