Key Facts
- A clear still frame does not prove a caption stays clear during motion.
- Check labels and spoken captions together.
- Use the actual words, including the longest line.
- Record collisions by shot and timestamp.
Give Every Piece Of Text A Job
Use captions for the spoken instruction and labels for details the viewer needs to locate. A heading should add something neither already says. When three text elements repeat the same message, remove the duplication before reducing the font size.
In this example, a Short shows how to fold a fabric napkin into a pocket. The spoken instruction is “Tuck the lower corner underneath.” A label pointing to the lower corner may help briefly. A large heading saying “TUCK IT” adds little while occupying the same useful space.
Trace the hands through the whole shot before choosing the reading area. In the napkin example, space that looks empty at the start becomes the work area during the fold. Reserve room for the full movement.
Build the framing plan first: https://dreamwild.ai/guides/vertical-video-composition/
Test The Export With The Longest Caption
Put the exported draft on a phone and watch it at a comfortable distance. Include the actual captions, especially the longest instruction. A monitor-sized preview can make text look generous even when it becomes cramped or hard to follow on the intended screen.
Example review: find the caption collisions in the napkin demonstration and choose a repair for each.
Check each moment where captions, labels, hands or player controls compete for space. Note the collision, make one clear revision and replay that moment in motion. The record below shows the decisions made in the fictional napkin example.
Scroll this table sideways to see every column.
| Review Moment | Collision Found | Revision | Retest |
|---|---|---|---|
| Fold begins | Caption covers the corner being lifted | Move the reading area above the hands | Corner remains visible throughout lift |
| Longest instruction | Last word touches the edge after line wrapping | Split at a natural phrase boundary | Both phrases readable without pausing |
| Object label appears | Arrow crosses the spoken caption | Show the label before the next caption phrase | Reading order is clear |
| Result held | Viewing controls overlap the finishing label | Move the label beside the pocket | Label clear on the tested surface |
| Final playback | Caption jumps between three positions | Keep one position for the folding sequence | Eye movement feels predictable |
Record the device, player and exact video version with the result. That makes the check usable after a handoff. If the crop or captions change later, you can identify which earlier observations need testing again.
Fix Wording And Phrasing Before Shrinking Text
Shorten the wording without changing the instruction. “Fold the bottom corner up” may communicate the same action more directly than “You now want to take the corner at the bottom and fold it upwards.”
Next, split the instruction at natural phrase boundaries. Keep a term together where possible so the viewer can read it once and return to the hands. A caption that changes halfway through an important phrase makes reading compete with following the fold.
Then adjust placement and contrast. A modest solid backing can keep the words legible as the scene moves, provided you leave deliberate space for it. Check what the backing hides. Shrinking everything should not be the automatic solution to a crowded frame.
Turn the script into caption phrases before detailed placement: https://dreamwild.ai/guides/spoken-words-to-captions/
Check The Player As Well As The Video
Use the intended player first, then a contrasting screen size if one is available. Watch with controls visible and during unobstructed playback, at ordinary brightness and viewing distance. Those checks are more useful than a guessed margin claimed to work on every phone.
Record which devices and viewing sizes you checked. If a second device is unavailable, note that limit in your production record and use a conservative layout with fewer edge-dependent details.
Watch once with sound muted to check whether the captions give enough time to follow the instruction. Then restore sound and listen for phrases that arrive too early or late. Readability and timing need to work together.
Approve The Action And Text In One Pass
The layout is ready when the viewer can read the instruction, identify its target and see the required action without pausing. For a dense diagram, finish the caption before asking them to inspect the detail. Give their attention a clear sequence.
Save a useful caption position as a starting point for similar scenes. Recheck it whenever the subject, crop or player changes: the next folding shot may put the hands exactly where the last video's captions sat.
Frequently Asked Questions
Is There One Safe Margin For Every Phone?
There is no universal margin prescribed here. Test the actual export on the viewing surfaces you intend to use, including the space occupied by visible controls.
Should Captions Follow A Moving Person?
Let captions move when doing so helps comprehension. Otherwise, a stable reading area gives viewers a predictable place to look and is easier to check for collisions.
Can I Put Labels And Captions On Screen Together?
Yes, if they do different jobs and remain readable together. Stagger or shorten them when the viewer has to choose which text to read.
Do I Need To Test The Export Again?
Yes. Check the file viewers will receive, including its crop, caption timing and text size. A correct editor preview does not establish that the export is correct.
Put the idea to work.
Build your scenes and captions together in DreamWild, then take the exported draft through the phone checks above.
Explore the creative studio https://dreamwild.ai/creative-studio/