You add subtitles by matching a timed text file to your video in an editor or player, or by letting a speech-recognition tool generate the timing for you. The first route gives frame-accurate control and works offline; the second cuts hours of manual typing but needs a correction pass. Which one fits depends on your deadline, your language, and whether the finished file will be uploaded, broadcast, or handed to a client.
This guide covers the formats that actually matter, the programs and services that accept them, and the failure points β dropped non-Latin characters, desynced timestamps, 32-character line limits β that catch people after export. Every method below is stable software behavior, not a speculative claim about a feature shipped last week.
Know Your Subtitle Formats Before Touching a Timeline
Two file families dominate: SRT (SubRip) and the newer WebVTT. SRT is plain text with numbered cues, a time range, and one or more lines of dialogue. WebVTT is the W3C-published format browsers expect for HTML5 elements and adds features SRT lacks β cue settings like line position and text alignment. If you're publishing to a website, WebVTT is the required format; if you're uploading to a video platform or handing off to a broadcast house, SRT remains the safe default.
| Format | Extension | Platform expectation | Styling support | Timing precision |
|---|---|---|---|---|
| SubRip | .srt | Editors, most video platforms, broadcast workflows | Basic (bold/italic tags) | Milliseconds |
| WebVTT | .vtt | HTML5 video, browsers, some streaming tools | Cue settings, CSS-adjacent | Milliseconds |
| ASS/SSA | .ass / .ssa | Anime fansubbing, advanced stylists | Full positioning, fonts, karaoke | Centiseconds in source |
| TTML/DFXP | .ttml / .dfxp | Broadcast, Netflix-style delivery specs | Extensive | Frame-level in some profiles |
| SCC | .scc | US broadcast (CEA-608/708 captions) | Limited to caption standards | Frame-accurate |
The distinction that trips people up: subtitles translate dialogue, while captions include sound effects and speaker labels for deaf and hard-of-hearing viewers. A file with only dialogue is a subtitle file even if it's technically an SRT β the format doesn't decide its purpose.
Encoding matters as much as format. SRT files saved as UTF-8 with BOM (byte order mark) can show a stray invisible character at the start of the first cue in some players; UTF-8 without BOM is the cleaner choice. If you're writing Arabic, Japanese, Korean, or any accented Latin text, saving as ANSI or Windows-1252 will corrupt characters on export even though they looked correct in the editor. Check the save dialog, not the preview.
Method 1: Type Timings Manually in Subtitle Edit (Free, Windows)
Subtitle Edit is the reference free tool for Windows users who want full control. Download it from the official project site, install, and use File > Open to load your video. The waveform view at the bottom of the window shows the audio, which is how you find the start and end of each line without scrubbing the whole file. Set the video position with the player, then click "Insert new subtitle at video position" (the keyboard shortcut by default is a bracket key, confirm it under Options > Settings > Shortcuts because it varies by version).
1: Open your video with File > Open and wait for the waveform to generate. 2: Play to the first spoken line and press the insert-subtitle shortcut to create a blank cue at that timestamp. 3: Type the line in the text box, then set the end time by pressing the same shortcut again when the line finishes. 4: Repeat to the end, then use Tools > Fix common errors to catch overlapping cues and too-short durations. 5: Save as .srt with File > Save as and pick UTF-8 encoding.
Failure handling: if the waveform never draws, the tool may be missing a codec β install the K-Lite codec pack or convert the audio to WAV first with an external tool and load that. If your saved file plays with garbled characters, you exported with the wrong encoding; reopen the SRT and re-save with UTF-8.
Method 2: Use the Platform's Built-In Caption Editor
Most major video platforms ship a caption editor that skips desktop software entirely. In YouTube Studio, open the video, go to Subtitles in the left menu, pick a language, and choose Add. From there you can auto-generate, upload a file, or type manually with the "Transcribe and set timings" workflow. That last option lets you paste a full transcript and have the platform distribute it across the timeline, then adjust each cue's start and end by dragging.
The advantage is native integration: caption edits sync to the video automatically, and the platform validates on save. The cost is less precision than a dedicated tool. For lecture-style content and interviews with clear speech, the tradeoff is fine.
Editing limits you'll hit: some platforms cap cue duration and character count per line, and won't warn you until you try to save. If a save fails silently, open the cue list and look for a warning icon next to the offending line rather than assuming the whole file is broken.
Method 3: Upload a Ready-Made SRT to Any Editor
If a translator or captioning vendor hands you an SRT, most editors will load it directly without retyping. In Adobe Premiere Pro, drag the SRT into the project panel and place it on a sequence above your footage; the captions appear as a stream you can style from the Essential Graphics panel. In DaVinci Resolve (version 18 and later), right-click in the Media Pool, choose Import > Subtitle, and select your file to create a subtitle track.
Format mismatches cause most import failures. A WebVTT file renamed to .srt will usually parse, but styling metadata in the WebVTT header gets discarded. ASS files with complex positioning may import as plain text, losing layout.
Watch for frame rate drift. An SRT timed against 23.976 fps footage will be about 4 seconds early by the end of an hour if the sequence runs at 24 fps β 0.1% per second doesn't sound like much until it accumulates. Confirm your sequence frame rate matches the source before importing any subtitle file, or retime afterward with the editor's offset control.
Method 4: Auto-Generate Captions, Then Correct Them
Speech recognition has shifted from unreliable novelty to a genuine time-saver for clear English dialogue, and the platforms provide it free. YouTube's auto-generation, Google's and Microsoft's speech-to-text APIs, and the transcription features in editors like Resolve all produce a first pass. The trap is treating the output as finished. Recognition accuracy on studio audio with a single speaker sits in the high 90s; on music beds, crosstalk, or accents outside the model's training mix, it drops sharply, and every misheard word is a listener complaint waiting.
Budget a correction pass of roughly one minute per minute of footage. That estimate assumes the transcript is mostly right and you're reading along with the audio; if the auto-caption is badly wrong, correcting it can cost more than transcribing from scratch by hand. Listen on headphones β speakers hide sibilance and trailing syllables the microphone caught.
Use the editor's search-and-replace for consistent mistranscriptions before line-by-line reading: character names and technical jargon repeat, so fixing them once globally saves dozens of individual edits. Then read once more with the sound on to catch the errors spellcheck can't see, like "defect" and "the fact".
Method 5: Generate Subtitles in the Video Editor Itself
NLEs now bundle AI transcription, so you can skip the external tool. In Adobe Premiere Pro (version 23.1 and later), use the Text panel's Transcribe Sequence feature to generate a transcript from the sequence audio, then convert it to a caption track. In DaVinci Resolve 18 and later, create a subtitle track from the Media Pool (Create Subtitle Track), right-click it, and use Auto Caption to transcribe the sequence. In Final Cut Pro (version 10.6.6 and later), select your clip and use the Create Captions command with auto-transcription.
Accuracy depends on the speech model, not your edit. Clean up audio first with a noise-reduction pass, because a loud HVAC hum will send the recognizer hunting for words in the drone. After generation, the caption track is editable cue-by-cue exactly like a manual import, so the workflow is generate, then correct, then export.
Export paths differ: Premiere saves via the Captions tab in the Export Settings dialog with a format dropdown (SRT and SCC among the options); Resolve exports through Deliver, ticking the subtitle export box in the render settings. Confirm the output includes every cue before you hand the file to anyone.
Method 6: Format by Hand in Notepad for Full Control
Writing an SRT in a plain text editor sounds archaic until you're on a machine without your usual software or need to patch one cue without opening a timeline. The structure is three lines per cue: a sequential number, a time range in hours:minutes:seconds,milliseconds --> hours:minutes:seconds,milliseconds format, then the text, followed by a blank line before the next cue. Note the comma before the milliseconds in SRT β WebVTT uses a period instead, and mixing the two breaks parsing in strict tools.
1: Open Notepad or any plain text editor and type cue 1: 1 on the first line. 2: On line two, enter the start time, space, -->, space, and end time, for example 00:00:03,400 --> 00:00:06,200. 3: On line three, type the dialogue, then leave one blank line before cue 2 starts. 4: Save as name.srt with UTF-8 encoding, choosing "All files" in the save-type dropdown so the editor doesn't append .txt. 5: Test in a player before delivery β load the video and SRT side by side in a tool like VLC with the subtitle track selected.
Failure handling: if the subtitles don't appear in the player, the most common cause is the double extension β name.srt.txt β from a save dialog that hid the real format. A second common cause is a mistyped time range missing the milliseconds, which makes the parser skip that cue silently. Open the file and check the first three cues character by character.
Method 7: Pick an Online Captioning Service for High-Volume Work
Agencies and studios handling dozens of videos per week usually route subtitling through a service rather than individual editors. Two publishing requirements drive the choice more than price: the accuracy ceiling you need, and your delivery format.
| Route | Speed on 1-hour video | Typical cost model | Accuracy ceiling | Best fit |
|---|---|---|---|---|
| Manual in Subtitle Edit | 4β8 hours | Your time | 99%+ with editing | Projects needing exact timing |
| Platform auto-captions | Minutes to generate, 1 hour to correct | Included with the platform | High 90s on clean audio | Social and web uploads |
| Editor AI transcription | Minutes to generate, 1 hour to correct | Bundled with the NLE license | Similar to platform auto | Workflows inside an NLE |
| Professional captioning service | Hours to a day, human-checked | Per minute, rate confirmed by the vendor | 99%+ with a human pass | Broadcast and accessibility compliance |
Confirm the service accepts your delivery format before signing β many quote SRT while broadcast delivery needs SCC, and the conversion step can add cost. Also ask whether the vendor does verbatim or edited captioning; verbatim transcripts include filler words that some broadcast standards reject.
If you're weighing an AI service against a human vendor, the deciding figure isn't the per-minute rate but your defect budget. For a two-minute social clip, one wrong word is the whole comment feed. For a 90-minute training video, a few errors are invisible.
Style Rules That Make Subtitles Readable
Readability lives in numbers. A widely used guideline set β the BBC's subtitle guidelines, published for its accessibility teams β caps each line at around 37 characters and each cue at two lines, with a reading speed ceiling near 160β180 words per minute. Your caption editor may default differently; check its settings rather than trusting the default. Too-long lines force the eye to chase text across the frame while the picture keeps moving.
Keep cues on screen for a minimum of about 1 second and a maximum of roughly 6 seconds, splitting longer dialogue into multiple cues. Break lines at natural syntactic boundaries, not mid-phrase β never put a determiner at the end of line one and its noun at the start of line two. Use italics for voice-over and off-screen speakers only when your delivery format supports styled tags; SRT does, WebVTT does via tags. Place the subtitle in the lower third but raise it above burned-in graphics like lower-thirds banners by using the editor's position control.
Reading speed matters more than character count once you're eyeballing cues. If a viewer can't read a line before it clears, the cue is too fast no matter how short it is. Time against your own reading rate during the correction pass.
Solve the Problems That Happen After Export
The subtitle is done when it plays correctly in the delivery environment β not when it looks right in the editor. Test on the player or platform the audience will use. Burned-in subtitles (rendered into the picture, called open captions) can't be switched off and survive any player, but they're permanent; a subtitle track (closed captions, delivered as a separate file or a track embedded in the container) can be toggled and edited later.
| Symptom | Likely cause | Fix |
|---|---|---|
| Garbled characters on non-Latin text | Saved in a legacy encoding (ANSI, Windows-1252) | Re-export as UTF-8 without BOM |
| Cues drift progressively | Frame rate mismatch between subtitle file and video | Retime with an offset, or re-encode to matching fps |
| Subtitles missing in the player | Wrong file path, wrong extension (double), or track not selected | Rename with a single extension and select the track manually |
| Overlapping cues rejected by a validator | End time of one cue is after the start time of the next | Use the editor's fix-common-errors pass |
| Vertical video lines cut off | Subtitle positioned outside the safe area | Enable safe margins and reposition per aspect ratio |
A quirk worth knowing: some video containers (MP4/MOV) can carry subtitle tracks embedded, while others (like older AVI workflows) leave captions as sidecar files. If a platform ignores your embedded track, deliver the SRT as a separate file β most platforms accept the sidecar.
Work Through a Full Plan for One Video
Start by identifying the purpose. A marketing clip uploaded to a platform can ship with auto-generated captions corrected once. A client deliverable or a broadcast master needs manual timing and a validation pass. The distinction sets how many hours go into the job before you even open the software.
1: Confirm the video's frame rate and aspect ratio from the source file's properties. 2: Decide delivery format β SRT for near-universal compatibility, WebVTT for web embedding, SCC only if a broadcast spec demands it. 3: Transcribe (auto-generate or type), then correct against the audio on headphones. 4: Set timings and split long cues; check reading speed and line breaks. 5: Export in UTF-8 and validate in the target player. 6: Test the file on the actual delivery platform before handover.
That sixth step catches more defects than the first five combined, because platforms apply their own validation and their own rendering defaults. A cue that looks clean in VLC can sit too low or too high on a phone screen. Load the platform's preview after upload and scan every cue.
FAQ
How long does it take to add subtitles to a one-hour video?
By hand in a dedicated editor, expect four to eight hours for transcription plus timing, depending on how clean the audio is. Auto-generation cuts the first pass to minutes, but the correction pass still runs close to one minute per minute of footage, so about an hour of correction on top. Professional services turn it around faster because multiple people or models split the work.
Can I add subtitles to a video without transcription software?
Yes. Write cues by hand in a plain text editor using the SRT three-line structure and save as .srt with UTF-8 encoding. The main traps are the double extension (file.srt.txt) and a time range missing its milliseconds, both of which make players skip the file silently. A player like VLC will load the SRT automatically if its filename matches the video's.
Why do my subtitles drift out of sync by the end of the video?
Almost always a frame rate mismatch. A subtitle file timed at 23.976 fps played against a 24 fps sequence runs about 0.1 percent fast, which accumulates over long runtimes. Fix it by matching the sequence frame rate before importing, or apply a timing offset in the editor. Confirm the source's true frame rate from the file properties, not from what the template says.
Do auto-generated captions meet accessibility standards?
Usually not on their own. Speech recognition misses crosstalk, background music cues, and non-speech audio that accessibility standards require. Auto-captioning is a starting point that needs a human correction pass, sound-effect references where dialogue alone is insufficient, and speaker identification in multi-person footage. Check the standard your jurisdiction follows before assuming a raw transcript qualifies.
What's the difference between burned-in subtitles and a subtitle track?
Burned-in (open) subtitles are rendered into the video frames and can't be turned off or edited without re-rendering the video. A subtitle track (closed captions) travels as a separate file or embedded stream the viewer can toggle. A track is easier to correct and translate later; burned-in text is guaranteed to display everywhere, which matters for old players and some social uploads.
How do I format subtitles for vertical video?
Vertical videos crop horizontal footage, so host the caption higher in the frame than the default lower-third, using the editor's position control rather than relying on the player's default placement. Confirm the caption sits inside the platform's safe area for phone screens, since edge controls can overlap text. Test on an actual phone before upload.
Is there a free subtitle editor better than typing in Notepad?
Subtitle Edit is free and widely used for manual timing on Windows; get it from the project's official site rather than a download aggregator. It handles the waveform, cue timing, and validation that a text editor can't, and it's fast once the shortcuts are set. On macOS, dedicated subtitle apps exist, but check the version's export format support before committing a project to one.
What happens if I upload an SRT that fails validation?
Most platforms reject the file or silently skip the failing cues rather than rejecting the whole upload. The cause is usually an unparseable time range or a cue whose end precedes its start. Open the file, check the first few cues against the three-line SRT structure, and re-save with the correct encoding. If the platform still refuses it, try importing into the platform's caption editor and letting it re-export in its own format.