Words and graphics you control (titles on slides, a news chyron, lower thirds,
captions, a headline that builds on a word) are composited over the raw footage with
hn4a-compositor, the app's compositor, so they stay separate from the picture: you
can change, re-time and re-render them without touching the footage. Don't generate
them as part of an image or video. The compositor plays a JSON layer spec live in a
browser page (a review plays the same .json live) and renders it frame-exact to MP4
with the spec's audio.
First question: what should the graphics help the viewer understand or notice?
Choose that job first, then the presentation and techniques. Several jobs can share a
shot; a camera move or a perspective plane supports them rather than competing with
an ad title or product explanation. Every leaf writes the same spec: its mechanics,
checks and live-spec gate are hn4a-compositor's.
Routes
| Purpose | Pick it when | Cost and time | What you get | Load |
|---|---|---|---|---|
| Emphasize a message | A benefit, claim, hook or action needs visual weight | Local; scope follows the layout | A readable hierarchy in the brand's type and colours | hn4a-compositor-motion-graphics; hn4a-compositor-brand-ad-titles for its ad title kit |
| Explain a product or process | The viewer needs to follow steps, relationships or a change | Local; a cut-up screen took under an hour, a rebuild a few hours | Diagrams, lists, screens and cues that clarify the idea | hn4a-compositor-motion-graphics; guide-show-a-product-on-screen for product evidence |
| Identify a person or context | The viewer needs a name, role, location or source | Local | A lower third or context label with time to read | hn4a-compositor-motion-graphics; hn4a-compositor-chyron for broadcast straps and tickers |
| Help the viewer read speech | Spoken content needs readable words on screen | Local | Captions timed to the actual track | guide-caption-a-video, then its selected leaf |
| Carry a song's words | A music or lyric video: the sung lines are part of the show | Local; the lyric times and the bar grid come first (guide-time-a-video-to-a-song) | Words that land on the sung syllables, pulse on the beat and take each scene's look | hn4a-compositor-kinetic-lyrics |
| Make graphics interact with the scene | A gesture, location or physical relationship carries the idea | Local; tracking or a matte adds preparation | Type that sits in the set or reacts to a person | hn4a-compositor-in-scene-titles, hn4a-compositor-hand-interactive-titles |
Supporting techniques
Use only the techniques that help the chosen purpose:
- Layout and brand title patterns:
hn4a-compositor-motion-graphicsfor general layouts;hn4a-compositor-brand-ad-titlesfor its hook, benefit and end-card helpers. - Sound on screen:
hn4a-compositor-audio-waveformdraws a file's real waveform (an audiogram, a sound effect's bars). - Depth and interaction:
hn4a-compositor-in-scene-titlesfor scene effects;hn4a-compositor-hand-interactive-titleswhen a gesture motivates the move;hn4a-compositor-perspective-planeswhen a graphic belongs on a surface. Use a matte (python-track-face-and-handsorbirefnet-matte, Choosing below) where a person should pass in front, and check title occlusion (python-check-title-occlusion). - Camera: choose pace from the tone and reading load.
hn4a-compositor-fast-ad-camerasupplies rhythmic pushes and whips;hn4a-compositor-calm-explainer-camerasupplies measured moves and holds. Keep the frame still when that best supports inspection. The video's category alone does not select its camera.
Product evidence and storyboards
guide-show-a-product-on-screen chooses a held screen, a recording, a staged reveal
(hn4a-compositor-make-product-ui-exciting) or a rebuild
(hn4a-compositor-rebuild-product-screen) from what the viewer must understand. Its
choice owns the product-screen routes; the graphics around the evidence follow the purposes above.
When the workflow needs a whole-piece storyboard before substantial production,
hn4a-compositor-storyboard proves its beats with cheap parts. Place it at that
workflow's existing storyboard gate, combining the first-look review when it answers
the same questions. A storyboard is a planning aid, not another kind of on-screen
message.
Text baked into the picture (drawn into a still, then animated) is a different
thing: it becomes part of the scene. Text that looks superimposed on the image usually
survives the motion, holding still while the background moves, but you can't change,
re-time or re-style it without re-rendering the shot. Use it only for text that belongs
to the world (a sign on the set, a logo on a prop: guide-design-a-set,
guide-keep-small-details), never for titles, chyrons or captions.
Cues
Graphics change on the words. The first look runs on word times estimated from the
script (Process, step 1); everything after it waits for the real inputs. Lock them
before the full build: the voice track, its word timings
(elevenlabs-speech-to-text, read with python-read-word-timings; track.json gives
each line's span), the footage, and a matte where needed.
- Start an entrance a little before its word: graphics in the source clips landed about 0.1 s early.
- Find shots at the finished-cut threshold (
ffmpeg-scene-cuts, Which threshold for which job), dissolves included. Find visual events (a screen tearing, a phone picked up) on a 2 fps contact sheet of the shot. - After a voice re-take, re-cue from its new transcript; never nudge numbers by hand.
Process (each gate a review)
- The first look. At the workflow's look step, show the opening scene or the
graphic relationship that needs proving (
hn4a-compositor, The first look). Use existing assets or estimated cues before making the full layout; develop a supplied direction rather than reopen it. Gate: the look, combined with the workflow's existing design review. - One spec per layout in the job folder (
slides/<name>.json), in the picked look, boxes laid out on the 1920×1080 canvas against a still of the reference slide, entrances from the cues. - A still per scene, on one sheet, looked at once for overflowing text, a missing font, a head covering a word. That catches mistakes; it doesn't approve the look. Save the still per cue and the frame-by-frame checks for the render.
- Gate on the whole design, live (
hn4a-compositor, Live spec, then render). The person watches the design move, not a sheet of stills: a caption style agreed in words ("bold slab captions") is not one they have seen, so get the yes before any render, though the render is free. - Render the approved spec and run its checks.
- Revisions re-render only the span that changed (
--into); a change to the look goes back to the person live first.
Choosing
- A person cut out over a background or over titles needs a real matte, never a colour
key. Free and local:
python-track-face-and-hands, a MediaPipe person matte with the face and hands tracked too, good on a webcam or talking head; orpython-matte-a-talking-head(Robust Video Matting) for a presenter cutout. Paid, when edges or furniture matter (hair on a busy room, a podium or chair that must stay):birefnet-matte, "General Use (Heavy)". Snap the matte's span to the shot's cut times. Otherwise keep the presenter in a window or bubble where its own background shows. - Type behind people: place it so a head or shoulder covers one letter or part of one,
never a whole word. Check the still and nudge the box, then measure every frame
(
python-check-title-occlusion). - A whole layout swap reads as one move when the old layer fades out over the new one fading in (a 0.5 s overlap), or as one box-and-crop keyframe move (full frame → window → bubble).
- Brand type and colour: local font files embed in the page; name the brand colour in the spec (Brand Sans, Globex Lime #B8E62E on #1B2E24).
Avoid
- Colour-keying for a cutout: it eats skin and hair edges.
- Remotion or any build step: the spec and one runtime are the whole tool.
- Titles, chyrons or captions generated into the image or video: composite them.
- A whole word covered by a head.
- A
scene > 0.25pass for cue times: it misses dissolves. - Text boxes sized for short copy: they don't shrink text; check the longest line.
- Building the whole design before the person has seen any of it move: they wait for every voice and picture, then see one look they didn't choose.
Related
-
guide-caption-a-video: whether and how to show the spoken words. -
hn4a-compositor-chyron: broadcast straps and tickers for the identification route. -
guide-voice-a-script: the track and word timings this reads. -
guide-design-a-set: where in-scene type sits. -
workflow-slide-deck-presenter: a talking head in slide layouts, end to end.


