Subtitle safe areas: keep important visuals visible

Make room for both the words and the picture

To keep subtitles from covering important visuals, review the final video with captions enabled, identify the cues that hide useful information, and resolve each collision in a tool or player that supports the required layout. Then check the delivered version on a small screen. A caption can fit comfortably inside the frame and still cover the very detail the viewer needs to see.

This guide is for recordings and subtitle files you own or have permission to edit and distribute. It separates three decisions: keeping text away from the edge, keeping essential picture information visible, and checking what the destination actually supports. W3C’s prerecorded-caption guidance says captions should leave relevant visual information unobstructed. A safe-area overlay alone cannot establish that result.

Separate frame margins from visual collisions

A subtitle safe area is a planning boundary for where captions should remain visible. Use the receiving broadcaster, platform or production specification when it supplies one. Do not treat a percentage copied from a different delivery workflow as a universal rule for every web player, phone and video shape.

An obstruction review asks a different question: what does the caption cover during its entire display interval? Look for a name card, a diagram label, a facial expression, a hand demonstrating a movement or a changing instrument reading. A subtitle placed well inside every margin may still hide any of these.

DCMP’s Captioning Key treats margins and placement as separate considerations. Its placement guidance moves captions away from important picture content when the player permits it, while its margin guidance addresses clipped characters. Use those principles to identify problems; use the destination’s documented capabilities to choose a repair.

Find collisions across the whole cue

Prepare a working subtitle export and the exact final video edit. Watch once for meaning, then make a separate visual pass with captions on. For every collision, note the cue identifier, its start and end, the hidden information and a proposed action. This is a small manual review sheet, not an automatic Polyglot report.

Play through the complete cue, including any shot change. Pausing on its first frame can miss a label that appears a second later or a hand that moves behind the text. Check the next caption too: solving one collision can produce a distracting jump or move the following cue onto another graphic.

The fictional paper-flower demonstration below illustrates the difference. Its spoken instruction is correct, but a caption across the lower part of the picture hides the fold being demonstrated. The aim is to preserve both the instruction and the visible fold, without making the subtitle appear before or after the speech.

Review copy:
  paper-flower-final-03.mp4
Track: English

Cue 12 | 00:18.000–00:21.000
Words: Fold the lower petal
  inward.
Hidden: Lower fold and
  fingertips
Candidate: Upper placement,
  if supported
Check: Whole cue and next
  shot

Cue 27 | 00:46.000–00:49.000
Words: Keep this crease
  visible.
Hidden: Upper diagram label
Candidate: Retain lower
  placement
Check: Label and caption
  remain readable

Choose a repair the delivery path can preserve

If the destination supports cue positioning, move the affected caption to an unobstructed area and inspect it throughout the cue. Top placement is a candidate, not an automatic fix: a title, face or diagram may already occupy that space. Keep placement stable across a passage when possible instead of making captions follow every moving object.

If positioning cannot survive delivery and you control the video edit, consider moving the on-screen label or changing the composition in the video editor. Keep an unchanged master. If a new edit changes shot duration or audio timing, recheck the subtitle synchronization against that new version before continuing.

If neither the subtitle layout nor the picture can be changed, record the collision as unresolved and agree a different delivery arrangement with the recipient. Do not call the file ready because it uploads successfully. Shrinking captions until they are difficult to read or deleting necessary speech trades one access problem for another.

Keep text and timing changes justified by the recording. Do not shift a caption into silence solely to avoid a graphic or paraphrase away a quantity that the viewer needs. If the actual problem is excessive text density, use the separate line-break and readability review before testing placement again.

Know where SRT and WebVTT stop helping

A basic SRT handoff is not a dependable way to communicate a custom screen position across players. For example, YouTube documents support for basic UTF-8 SRT without styling markup, while its WebVTT entry explicitly supports positioning with limited styling. That is a YouTube-specific distinction; another destination needs its own format check.

The W3C WebVTT Candidate Recommendation Draft defines cue settings for layout, including line, position and size. The existence of those settings does not prove that a particular editor preserves them or that a receiving platform renders them as intended. Confirm both the file contents and the actual playback result.

Polyglot can import, review and export timed text, but its WebVTT import deliberately removes presentation settings and reports that removal. It does not preserve an imported positioned layout through a round trip. Keep the original positioned file separately. Finish text and translation review before applying destination-specific positioning in a suitable authoring tool, and recheck the final export after any later content revision.

Changing an extension from .srt to .vtt does not add placement. Likewise, seeing a caption above the picture in one local preview does not establish where a different player will place it. Treat the final player as a required acceptance check.

Review each language and viewing layout

Repeat the visual pass for every delivered language. A translation may wrap into more lines and cover more of the picture even when its timestamps match the original. A combined bilingual review track needs its own check; it is not evidence that either separate public language track has passed.

Check the intended embedded player and full-screen view, plus a narrow phone-sized view. Where the destination allows caption-size changes, try a larger available setting and note the result. Show and hide playback controls, then seek back into the problem passage. These checks help reveal collisions with controls, reflow or a caption box that differs from the one in the authoring preview.

If you distribute a vertical or square edit as well as a landscape video, review each finished edit separately. Cropping changes which picture details sit behind captions. A successful landscape check cannot approve an uninspected vertical version.

Save a screenshot or timecoded note for each repaired passage, naming the file, language and player used. A still is useful evidence of placement at that moment; it does not replace playback across the complete cue.

Hand off the result with its remaining limits

Deliver the reviewed subtitle file with the matching video revision and a short placement note. Keep the plain text review export, the final positioned deliverable when applicable, and the original received file distinguishable. A recipient should be able to tell which file was actually checked and which software stage added positioning.

For the paper-flower example, acceptance means the viewer can read the instruction while seeing the fold for all of cue 12, and still see the upper diagram during cue 27. It also means the chosen positioning survives the recipient’s player. If only a local preview was checked, leave destination approval pending.

The checklist below is a suggested handoff record. Completing it documents the checks you performed; it is not a certification of accessibility or proof that every possible device will behave identically.

Video revision:
  __________________
Track / language:
  ________________
Delivered file:
  __________________
Player and view checked:
  _________

[ ] Frame edges do not clip
  text
[ ] Essential visuals stay
  visible
[ ] Full cue intervals
  reviewed
[ ] Every language reviewed
[ ] Narrow view checked
[ ] Final file preserves
  placement

Unresolved collisions:
  ___________

Keep exploring.