20 Common Video Subtitle Questions

When making subtitles, problems cluster around a few fixed stages: how to write the format, how speech gets recognized, how translation stays error-free, and how the final cut gets made. The 20 most common questions, grouped into those four — come here first when stuck, and you can usually locate the issue directly. Not exhaustive; just the highest-frequency, most beginner-tripping ones.

The subtitle pipeline from production to final cut
Subtitle processing runs through recognition, translation, spec checking and final cut; each stage maps to a class of common problems, and troubleshooting by stage locates faster

One: format (5 questions)

What's the difference between SRT and VTT?

SRT is plain text with the simplest structure; VTT adds a WEBVTT header, supports styling and positioning, and mainly attaches to web videos. Convert with the SRT to VTT tool.

What are the parts of an SRT cue?

Three: index, timeline (00:00:01,000 --> 00:00:02,500), text, with a blank line between cues.

What format is the timeline?

SRT uses a comma for milliseconds (00:00:01,000); VTT uses a dot (00:00:01.000). Don't mix them.

What encoding should the subtitle file use?

UTF-8 without BOM, LF line endings — most stable across platforms, and it avoids boxes and question marks.

How does ASS differ from SRT?

ASS supports fonts, colors, positioning and other styles, suited to precise burn-in; SRT is text only. Convert with the SRT to ASS tool.

Two: recognition (5 questions)

Which languages does speech recognition support?

Source languages: Chinese, English, Japanese, Korean; recognized into source subtitles.

What if recognition is inaccurate?

Proof by ear cue by cue, focusing on proper nouns and numbers; then run the subtitle cleaner for overlaps and format.

Can noisy videos be recognized?

Yes, but noise lowers accuracy. Quiet recording environments and close miking give steadier results.

Are uploaded videos saved?

No. Original videos uploaded to SmileSub are deleted when processing completes — no long-term storage.

Can recognition output be used as subtitles directly?

Proofread before publishing, especially terminology and numbers, so typos don't reach the final cut.

Three: translation (5 questions)

Which languages does translation support?

20 languages including Chinese, English, Japanese and Korean, translating among each other.

Does translation scramble the timeline?

Professional tools don't: the timeline is parsed separately, only the text is translated, then reassembled on the original axis.

Can machine translation be published directly?

Informational voiceover, yes; professional terminology and brand voice deserve a human pass.

How do I keep terminology consistent?

Fix a glossary before translating — one concept, one rendering, throughout.

How do I make bilingual subtitles?

The bilingual SRT generator — upload an SRT and get original plus translation directly.

Four: final cut (5 questions)

Soft or burned subtitles?

Platforms that take CC: soft, traded for SEO; social and ads: burned to guarantee reach. Both share one SRT.

Can subtitles be changed after burning?

No — burn-in is permanent. Changes go back to the source SRT and re-export, so proofread before burning.

What if subtitles overflow the picture?

Shorten lines or reduce font size; lay out against the safe area. The spec check lists every offending cue.

What reading speed is right?

Under 9 chars/second for Chinese, 20 for Latin, 1–7 seconds per cue.

How many subtitle versions does one video need?

Depends on channel: YouTube-style platforms take soft tracks, social gets burns, more markets get more language tracks.

For a fuller delivery trade-off, see burned-in vs soft subtitles; for the complete workflow, see the full subtitle translation workflow.

FAQ

Are uploaded videos stored by the platform?

No. SmileSub deletes the original video when processing completes — no long-term storage. That's different from tools that keep files on servers; sensitive material can be uploaded with confidence.

Does translation scramble the timeline?

It shouldn't. Translation acts on text only; the timeline is separately parsed data, reassembled on the original axis. What does scramble it is manual copy-paste — structured data becomes continuous text, and the re-splitting has no validation.

How do I fix subtitles overflowing the picture?

Shorten the per-line count first, then font size — and always respect the safe area. A spec-check tool lists every violating cue in one pass, faster than hunting by eye.