Tutorials · 工作流
You have local subtitles. Now get them into the finished video.
Generating subtitles is a middle step, not the end. Most people get stuck on the next one: the file exists, but the editor rejects it or the timing is off. This walks through what happens between a subtitle file and a finished video, in order.
1. Make sure the file is actually clean
A usable SRT satisfies three things: sequential numbering, timestamps in the form `00:00:01,000 --> 00:00:03,500` (a comma, not a period), and text free of stray HTML tags. MonoSub’s output meets this by default, but if another tool produced the file, open it and check — most "import failed" reports are format problems, not software problems.
- Numbering starts at 1 and never skips; gaps make some players drop cues
- SRT separates milliseconds with a comma, VTT with a period
- Lines over ~42 characters get wrapped by most players, but not at sensible points
- A blank line separates cues; an extra one can make the next cue disappear
2. Pick the right format
Different downstream tools want different containers. The usual symptom of picking wrong is not an error — it is "the subtitles just are not there".
| Downstream | Format | Note |
|---|---|---|
| Premiere Pro / DaVinci Resolve | SRT | Both import it natively as a caption track |
| Final Cut Pro | SRT | Via its caption workflow, or converted to .fcpxml |
| YouTube / Bilibili upload | SRT or VTT | YouTube takes both; VTT gives finer style control |
| Web players (video.js and friends) | VTT | A `<track>` element only accepts VTT |
| Plain text / further editing | TXT or MD | Drops the timeline, keeps the words |
| Scripted processing | JSON | Structured timestamps for code to consume |
3. Deal with timing drift
If the subtitles are uniformly early or late by a fixed amount, that is the easy case: every editor has a shift-all-times control — enter +2000 ms or −1500 ms and you are done.
If the start is aligned but the drift grows, that is a frame-rate mismatch — classically 23.976 fps footage interpreted as 25 fps. Shifting cannot fix it; the timeline has to be scaled, or the true frame rate confirmed before generating subtitles at all.
4. Sidecar or burned in
A sidecar subtitle is a separate file a player loads; burning in draws the text into the pixels. The trade-off is straightforward:
| Sidecar | Burned in | |
|---|---|---|
| Editable | ✓ Change text and timing any time | ✗ Re-export to change a word |
| Can be turned off | ✓ Viewer decides | ✗ Permanent |
| Platform support | Varies; some reject uploads | ✓ Plays everywhere |
| Style control | Limited by the player | ✓ Complete |
| Best for | Archives, multiple languages, accessibility | Short-form, social, ad creative |
MonoSub produces subtitle files and stops there; burn-in belongs to your editor or to ffmpeg. On the command line, `ffmpeg -i in.mp4 -vf subtitles=sub.srt out.mp4` is enough, at the cost of one re-encode.
What word-level timestamps buy you
Cue-level timestamps only say when a line begins. Word-level timestamps record the start and end of every word, which makes this possible:
- Karaoke-style highlighting, word by word — the most common short-form effect
- Editing video by editing text: delete a word and its audio goes with it
- Automatic cut points for silence removal, without listening through
- Re-segmenting a long line into two cues without breaking the timing
MonoSub writes word and silence data alongside the subtitles precisely for these uses. If your workflow is "upload the video with captions", cue-level is enough; word-level is required for per-word animation or text-based editing.
Proofreading is the expensive part
Every speech recogniser makes mistakes. The goal is not eliminating them but lowering the cost of fixing them. Two habits that work:
First, read the whole thing once and only mark errors — do not fix as you go, or you will lose the overview in the details. Second, keep a list of proper nouns and replace them in bulk; name, product and jargon errors are almost always repeated.
The verdict
Generating subtitles is the most automated step in the chain. What decides the finished result is the three things after it: the right format, a timeline that lines up, and whether you actually proofread. Software removes most of the pain from the first two. Nobody can remove the third for you.
MonoSub
Written by the people building a local subtitle tool. What it does and does not do is all in here.
Prices and features follow each vendor’s site; data was checked on the date noted in each post.
Free to download, Pro only for batch
Processing one file at a time is free. Batch work needs a one-time Pro: $9.99 / ¥48 as a Mac App Store in-app purchase, or ¥36 / $8.60 as a direct-store offline licence.
Related reading
MonoSub vs VideoCaptioner: free and open source, or a free app with a one-time Pro
27 Oct 2026 · 5 min read
ComparisonsMonoSub vs MacWhisper: two one-time-purchase local transcribers, compared
27 Oct 2026 · 5 min read
ComparisonsMonoSub vs cloud transcription: 讯飞听见, 通义听悟 and 飞书妙记
27 Oct 2026 · 5 min read
TipsThe biggest speech model is often the wrong choice
21 Oct 2026 · 4 min read