The term “caption” can mean two different things in a publishing workflow: the text accompanying a post, or the words displayed during the video. This guide concerns the second meaning.
The key points
- Burned-in subtitles form part of the video image; separate tracks depend on the player.
- Check wording, timing and readability before approving the export.
- Inspect the final file and any overlap with additional caption layers.
There is another distinction inside the video. Some captions are rendered into the image; others are supplied as a separate timed track. Knowing which version you have prevents surprises when moving a clip between tools or publishing destinations.
Choose the delivery format first
Choose how the subtitle text should be displayed.
Burned-in text
- Burned-in text is always visible in that video file.
A separate track
- A supported separate track can be shown or hidden independently.
The W3C caption guidance describes this difference between open and closed captions.
Neither choice removes the need for accurate text. Nor does selecting a visual style establish that the whole video is accessible. Consider what viewers need to understand, including relevant sounds and speaker changes.
| Your need | Useful approach to evaluate |
|---|---|
| A fixed visual treatment in the exported image | Burned-in captions |
| Viewer control over displaying a track | Separate captions where supported |
| Several language versions | Preserve editable text and a clean base |
| Later corrections | Keep the source project and reviewed transcript |
A practical archive includes the editable project, the approved text and the exported version used for publication. That makes a later correction traceable.

Subtitles in the picture, no extra app
Everything happens in the browser, right before scheduling.
- Creates subtitles, you fix every word, BestPost burns them in
- Trims in the browser and cuts silence and filler words
- Runs in the browser, on your computer and on your phone
Review meaning before appearance
Automatic transcription is a draft. Listen to the audio while reading it, with particular attention to names, measurements and negatives.
Consider an invented training example: “Do not tighten the clamp yet.” Losing “not” changes the instruction. A spellchecker may not identify that problem because both versions are grammatical.
Check words that sound alike and terminology specific to your business. When the audio is unclear, resolve the uncertainty with the speaker or source material. Do not replace it with a confident guess simply to finish the export.
Make the text work with the video
Review subtitles against these conditions:
- Timing: Review line breaks and timing at normal playback speed. Text should appear with the relevant speech, not several moments before the idea is introduced. It should also remain readable without forcing a viewer to pause every sentence.
- Background: Look at the busiest background, not only a clean opening frame. A style that works against a wall may fail during a close-up of a patterned product.
- Essential detail: Position the text around the information viewers need to see. For a cooking demonstration, a caption over the measuring marks can hide the very detail the narration explains. For a screen recording, avoid covering the control being selected.
The layout is then judged against the actual footage.
Inspect the export, not just the preview
Inspect the finished file in this order:
Open the export
Open the rendered video outside the editor. If you expect burned-in captions, they should be visible as part of the image without loading a separate text file.
Check timing
Watch transitions and the ending as well as the introduction. A revised cut can leave timing errors later in the clip even when the first sentence looks correct.
Inspect the destination
Then inspect the intended viewing surface. Social interfaces add controls and text around the video. A caption position that looks clear in a bare player can be covered in the destination app.
The review then covers both the file and its intended viewing context.
This is a visual check, not a universal pixel formula. Interface arrangements and display states can vary.
You correct, BestPost burns them in
Check word by word, then render them into the video.
- Every text stays editable until you post
- One style preset for the subtitles of every video
- Pulls the thumbnail straight from the video
What if two caption layers appear?
Check these points when caption layers overlap:
- Additional layer: A burned-in layer can coexist with a platform-generated or separate caption layer. Check whether the result obscures the video or becomes hard to read. Review the controls available in the actual publishing and viewing workflow.
- Viewer controls: Do not remove useful viewer options reflexively. First identify which text belongs to the video image and which belongs to the player. Then choose the intended combination for that destination.
- File versions: Use file names that distinguish versions with and without rendered text. This is particularly helpful when you produce several languages.
This identifies the intended combination for each destination.
Decide which deliverables the project needs
List the required versions before designing captions. You may need a rendered English version, another language and an editable base for future changes.
Give each deliverable a clear name and approval state. A later publisher should be able to identify the language and text treatment without guessing from a sequence of export dates.
Keep transcript approval and visual approval distinct. The same person can perform both, but they answer different questions. Correct wording can still be placed badly, and an attractive layout can still say the wrong thing.
For a client handoff, record which version was reviewed. A general “captions approved” message becomes ambiguous after several revisions.
Test a shared style on a demanding clip
Choose an example with longer phrases, changing backgrounds and important action in the frame. A style tested only against a clean wall is a weak reference for a visually busy series.
Apply the treatment to the remaining clips, then inspect each export. Product names, language changes and longer sentences can alter line breaks even when the style settings are identical.
Look beyond simple visibility. Does animated emphasis distract from the demonstrated step? Does a line appear before the viewer has enough context to understand it? These are editorial judgments that a preset cannot settle for every scene.
You can simplify a treatment without making the content less professional. The purpose is readable information synchronized with the video.
One style preset for all your videos
So your feed does not look different with every clip.
- Takes 100 videos and more in one go
- Sends the same video to Instagram, TikTok, YouTube, Facebook and X
- Free for 14 days, no credit card
Use a specific acceptance record
Record four outcomes: wording checked, timing checked, placement reviewed and correct file selected. This small checklist works across a batch while preserving individual accountability.
When a version fails, identify the scene and required correction. “Fix captions” leaves too much open. “Correct the product name in the second sentence and recheck its line break” is actionable.
After a change, inspect consequences as well as the original error. A longer corrected phrase can need different spacing, while an altered cut can affect later timing.
Keep the approval record with the asset. It should explain why this export, rather than a similar working file, is ready for the publication workflow.
Handle corrections as new versions
An error in burned-in text requires a corrected video export. Updating the transcript alone does not change an existing rendered file.
After a correction, inspect any scheduled entry that uses the earlier version. Confirm the file and text together. Keep the revised asset clearly identified so a colleague cannot accidentally reselect the previous export.
BestPost's subtitle feature supports preparing the text and video. Your final check should still confirm the exact version that will be published, including its wording, timing and visible placement.