The Short Answer
Use a separate timed text file, such as SRT, when the platform or buyer can read one, and burn the text into the picture when you need it to show everywhere, like a social trailer, a festival screening with open captions, or a pitch link. Most films need both: sidecar files for the delivery and a burned-in version for marketing and for festival screenings that require it.
The words get used loosely, so start with definitions.
Captions, Subtitles and SDH
- Subtitles translate or transcribe dialogue for viewers who can hear the audio. The W3C accessibility guidance describes subtitles as dialogue only, in a different language, without the important non-speech sounds.
- Captions include dialogue plus meaningful sounds, such as a door slam, music cues and who is speaking, for viewers who are deaf or hard of hearing.
- SDH (subtitles for the deaf and hard of hearing) is the captions style delivered as subtitles, often in the same language as the film.
- Closed means the viewer can turn the text on and off. Open means it is part of the picture and always visible. The W3C describes open captions as embedded in the video, so they are always visible and need no player support.
- Forced narrative subtitles show only the foreign-language lines in a film that is otherwise in the viewer's language.
SRT or Burned In: How to Choose
| Situation | Use | Why |
|---|---|---|
| Streaming or distributor delivery | Timed text file in the format named by the buyer | Buyers need files they can check, edit and translate |
| Social trailers and ads | Burned-in captions, plus a clean version | Many people watch without sound, and some platforms do not read sidecar files |
| Festival screening | Whatever the festival asks for | Some require open captions or an SDH version |
| Pitch or screener link | Burned-in, or a sidecar the player supports | Players differ, and open text always shows |
| Website video | WebVTT or SRT, depending on the player | Viewers can switch it on, and it supports accessibility |
Sundance's 2026 materials give a real example of a festival rule: English-language projects must deliver an open caption version of their DCP for in-person screenings, and non-English projects must deliver a version with standard English subtitles and one with English SDH subtitles. The festival also recommends an audio description track. Do not assume every festival asks for this; read each one.
Common File Types
- SRT is plain text with a number, a start and end time, and the text. It is simple and widely accepted, but has no styling or positioning of its own.
- WebVTT is the web standard for video players and allows some styling and positioning.
- Broadcast and streaming formats, such as TTML or iTT variants, SCC and others, are what many buyers actually name. The exact format depends on the buyer, so check the delivery specification before you make files.
- Burned-in text is part of the image and cannot be changed after export.
Timing and Style Rules
Each buyer publishes a style guide, and they differ by language. Netflix publishes detailed timed text style guides, and its English SDH guide specifies 42 characters per line, two lines at most, a minimum duration of five-sixths of a second per subtitle and a maximum of seven seconds, when we checked on 2026-10-17. These are one buyer's rules, shown as an example of the level of detail you may be asked to meet. Follow the guide you are given.
Useful habits that apply almost everywhere:
- Break lines at natural phrase boundaries, not in the middle of a name or verb phrase.
- Keep each caption on screen long enough to read, and do not flash text for a fraction of a second.
- Match timing to the speech, and begin captions when the speaker starts.
- Identify speakers when they are off screen or unclear.
- Describe sounds that matter to the story, in brackets.
- Use italics or a consistent mark for off-screen voices and narration if your guide says so.
- Keep text clear of the area where other important picture information sits.
Accessibility Basics
In the United States, the FCC's caption quality rules for television programming name four areas: accuracy, synchronicity, completeness and placement. Captions should convey the dialogue and sounds fully, line up with speech, run from start to finish, and avoid covering important on-screen content. Those rules apply to certain television programming, and whether a particular film or platform is covered is a question for a qualified accessibility or legal adviser. They still make a useful quality checklist.
For video on the web, WCAG expects captions for prerecorded audio in synchronized media, and W3C lists open captions as one technique to meet it.
Where Mistakes Creep In
- Auto-generated text is not finished captions. Speech-to-text is a fast first draft, but it often gets names, accents, technical words and punctuation wrong. Have a person review every line.
- Wrong frame rate. A caption file made at the wrong frame rate or timecode start drifts out of sync. Confirm the frame rate and start timecode of the master.
- Text after compression. A festival warns that subtitles can become indistinct once a site recompresses a screener, so test on the real link. Use a legible size and a clear outline or box.
- Burned-in text in the only master. Always keep a clean version without text.
- Unsafe positioning. Text too close to the edge disappears on some screens and vertical crops. See trailer social cutdown specs.
- Spelling inconsistencies. Character names in the credits, script and captions must match.
- Late changes. If picture changes after the captions are made, retime the files. A one-frame shift in the cut can break every line afterwards.
A Practical Workflow
- Lock picture, or accept that you will retime.
- Get a transcript or dialogue list.
- Create the captions with a tool or a vendor, then review by eye against the picture.
- Export the sidecar files in the format named by the buyer.
- Make a burned-in version for marketing and festivals.
- Check each deliverable against the delivery checklist in indie film post deliverables.
Want a person to help?
Our post-production team prepares caption, subtitle and SDH deliverables, burned-in versions for trailers and festival-ready exports, and checks them against the delivery spec. See our post-production services, or contact us with the destination and the spec sheet.
Prepare captions and subtitles for a film
Check the delivery spec
Find the required file formats, language versions and style guide from the buyer or festival.
Prepare a transcript
Create a dialogue list or transcript from the locked picture and check names against the credits.
Create and review the captions
Make the timed text, then review every line against the picture for accuracy, timing and line breaks.
Export the sidecar files
Export in the named format at the correct frame rate and start timecode.
Make a burned-in version
Render open captions for marketing or festival use inside the safe area and keep a clean master.
Test on the real destination
Play the files on the target platform or player and check legibility after compression.
Questions people ask
What is the difference between captions and subtitles?
Subtitles carry dialogue, often translated, for viewers who can hear. Captions also include meaningful sounds and speaker identification for viewers who are deaf or hard of hearing. SDH is subtitles written in the captions style.
Should I use an SRT file or burned-in captions?
Use a timed text file where the platform or buyer accepts it, and burned-in captions for social media, pitch links and festival screenings that need open captions. Keep a clean version without burned-in text.
What are open captions?
Open captions are part of the picture, so they always show and need no player support. Closed captions can be switched on and off by the viewer.
Are auto-generated captions good enough?
They are a useful first draft, but they often get names, accents and technical words wrong. Have a person review and correct them before delivery or publication.
Written by the team at Clouds Agency, a Los Angeles creative and production consulting agency.