Reaction, Duet and Stitch Formats (2026)
Reaction, duet and stitch are the formats where one video answers another. They run the comment layer of TikTok, Reels and Shorts — and Q4 feeds them constantly: costume reactions around 10⁄31, deal-checking duets through Black Friday on 11⁄27, gift-take stitches into 12⁄25, hot takes for 1⁄1.
They’re also the hardest to prompt: you’re describing two frames at once plus their relationship — which pane leads, which responds, when each moves. But 2026 models — Seedance 2.5, Kling 4.0, Veo 3, Sora 2 — can generate split-screen compositions in one clip, and when they can’t hold both panes, you generate each and composite.
But “split screen reaction video” is not a prompt: pane ratios, pane roles and beat mapping must be written explicitly or the halves wander apart. Here’s the grammar, seven shot elements, and three templates.
Three formats, three grammars
Reaction: one pane plays the source, the other shows the responder. Panes are usually unequal — a large source with a smaller face, or a top/bottom stack. The responder’s beats are mapped to the source’s: the point, the laugh, the lean-in all land on moments in the other pane.
Duet: both panes perform at once, equally framed. It lives on synchronization — mirrored moves, shared timing, poses that line up at the same timestamp. Equal weight, equal light, one clock.
Stitch: sequential, not simultaneous. The source runs first (3-5 seconds), then a hard cut to the take. The cut is the punchline — its timing matters more than what follows.
In real platform features, duet and stitch mean responding to someone’s actual post. In generated video, the “other pane” must be your own or licensed material — you can’t generate someone’s face or footage into your clip.
The seven elements of split-screen interaction
| Element | Interaction version | What to write in the prompt |
|---|---|---|
| Subject | Two panes, two roles, each defined separately | “Pane A (source): a generic man at a table doing a card trick; pane B (reactor): a woman in her 20s against a plain wall” |
| Framing | Layout and pane ratio written explicitly | “Hard vertical split at center, thin dark divider; left pane 50%, right pane 50%” |
| Camera | Each pane gets its own camera spec; usually both locked | “Both panes locked; pane B medium close-up, pane A medium wide” |
| Motion | Interaction beats: responses mapped onto source actions | “He reveals the card at 5s; she leans in at 5s, eyes widen at 6s, points at 7s” |
| Light & color | Panes need contrast, not competition | “Pane A cool light; pane B warm key — the panes read apart at a glance” |
| Timing | One shared clock across both panes | “Both panes live 0-15s; reaction beats at 6s, 7s, 9s; hold 12-15s” |
| Sound | Audio levels for each pane; reaction voice over source | “Card sounds low under her clear reaction line and laugh, soft room tone” |
The Timing row is the format. Two panes on one clock is the whole trick; separate clocks look like two videos glued together.
Prompting two panes: compose or composite
Compose in one prompt for a single generated clip: define both panes, the divider and one shared timeline in one block. Best when panes stay active together.
Composite in the edit when panes drift — the model blends the panes, changes outfits mid-clip, or forgets which pane is which. Generate pane A and pane B with matching framing lines and one shared clock, then stack them in the edit — also the only way to get a clean vertical divider.
Stitch is just a cut. Two clips in sequence: the source (3-5s) and your take, the cut written as a beat (“hard cut at 4s on her first word”).
Workflow: extract the interaction beats
- Find a clip where the panes feel connected. You want the one where the reaction felt timed, not the loudest laugh.
- Run it through SocialToPrompt. The frame-by-frame breakdown shows the beat map: which source moment triggered which response, pane sizes, when the cut landed. Videos up to 20s; new accounts get 3 credits, a full run costs 4 (1 extraction + 3 generations).
- Keep the beat map, swap both panes. Same trigger-and-response timing, same layout. Your characters, your source action, your palette.
- Test the divider. If the split looks soft or blended, switch to compositing two clips.
- Ride the Q4 conversation. 10⁄31 costume reactions, 11⁄27 deal duets, holiday gift stitches, 1⁄1 takes — interaction grammar stays constant, the hook changes.
Copy-paste templates
1. Side-by-side reaction (9:16, ~15s)
Vertical 9:16, reaction split screen, hard vertical split at center, thin dark divider, both panes locked.
Pane A left "source": a generic man in a plain hoodie doing a card trick at a table, plain wall, cool light,
medium wide. Pane B right "reactor": a woman in her 20s in a knit sweater against a plain warm wall, warm
key light, medium close-up. Motion: pane A riffles cards at 2s and reveals one at 5s; pane B watches with
small head movement, leans in at 5s, eyes widen at 6s, points at 7s, laughs at 9s. Timing: both panes live
0-15s, reaction beats at 6s, 7s and 9s, hold still 12-15s. Sound: card sounds low under her clear reaction
line and laugh, soft room tone. No on-screen text. Total 15 seconds.
2. Top/bottom dance duet (9:16, ~18s)
Vertical 9:16, duet split screen, hard horizontal split at mid-frame, thin light divider. Top pane "lead":
a generic dancer in a plain black outfit on a plain floor. Bottom pane "mirror": another generic dancer in
plain gray, framed identically, same cool-neutral lighting both panes, both locked wide shots. Motion: both
perform the same three-move phrase on one clock — arms up at 3s, spin at 7s, drop and pose at 11s — bottom pane mirrored so poses line up across the divider. Timing: phrase 0-14s, hold final pose 14-18s,
both panes live throughout. Sound: one shared beat track through the clip, soft footfalls under it. No
on-screen text, no logos. Total 18 seconds.
3. Stitch take (9:16, ~15s)
Vertical 9:16, stitch format, full frame, two sequential shots. Shot 1 (0-4s): a generic man in a plain tee
at a desk holds up a badly wrapped gift, plain room, soft daylight; he shrugs. Hard cut at 4s. Shot 2
(4-15s): a woman in her 30s in a cardigan, medium close-up against a plain warm wall, warm key light; she
talks to camera, holds up neat wrapping paper at 8s, demonstrates one fold at 10-13s. Camera: locked both
shots. Timing: source 0-4s, cut at 4s on her first word, gesture beats at 8s and 10s, hold 13-15s. Sound:
source room tone low, her clear close voice from 4s, soft tap as the paper is placed. No on-screen text.
Total 15 seconds.
Pre-flight checklist
- [ ] Format named: reaction, duet or stitch — each has a different structure
- [ ] Both panes defined: who is in each, what each does
- [ ] Pane layout and ratio written explicitly (50⁄50, 70⁄30, top/bottom)
- [ ] One shared clock: all beats given on the same timeline
- [ ] Trigger-and-response beats mapped (source action → reaction action)
- [ ] Panes contrast in light or wardrobe so they read apart instantly
- [ ] Stitch cut written as a beat with an exact second
- [ ] The other pane is your own or licensed material — no lifted faces or footage
Common mistakes and FAQ
Q: One prompt or two composited clips? Start with one prompt for simple layouts. If panes blend, drift or swap, generate separately and composite — clean dividers rarely come from a single generation.
Q: Can the other pane be a real creator’s video? In the platform’s native duet/stitch feature, that’s the point. In generated or republished content, no — keep generated panes original, and never re-create someone’s face or voice.
Q: How do I keep both panes in sync? Write one timeline, not two. “At 5s he reveals the card; at 5s she leans in” — same clock, both panes, every beat.
Find the split-screen clip where the reaction felt perfectly timed, run it through SocialToPrompt, and rebuild its beat map around your two panes.
Share Article