通过 Upwork 认识我? 如果您是通过 Upwork 找到我并希望合作,请通过 Upwork 联系我。Upwork 客户在 Upwork 之外发送的消息将不会得到回复。
对比🎬 字幕与隐藏字幕

Which caption deliverable to request: SCC, TTML, WebVTT or DCP

SCC for US broadcast, TTML/IMSC for streaming, WebVTT for web video, SMPTE ST 428-7 for DCP cinema: the format decision guide with QC step for each target.

Dao Huy (Lucas)8 min read8 部分 · 对照表
Which caption deliverable to request: SCC, TTML, WebVTT or DCP

💡 For US broadcast, request SCC (CEA-608 service, 29.97 drop-frame timecode) or SMPTE-TT for a native CEA-708 service. For streaming, ask for TTML/IMSC or SRT per the platform's spec. HTML5 web video uses WebVTT. DCP cinema requires SMPTE ST 428-7 subtitle XML. Each format carries different encoding, timing, and QC requirements - knowing this before the job saves a costly revision round.

Key takeaways

  • US broadcast 608: request SCC at 29.97 drop-frame timecode. For a native CEA-708 styling service: request SMPTE-TT (ST 2052-1 XML).
  • Streaming (Netflix and similar): request TTML (IMSC-compliant) or SRT depending on the platform's Timed Text Style Guide. Never assume one format fits all streamers.
  • HTML5 web video: request WebVTT (.vtt). It is the native HTML <track> format and the only format with native browser cue-positioning support.
  • DCP cinema: request SMPTE ST 428-7 subtitle XML. Interop XML is a legacy alternative accepted by older projectors, but SMPTE is the current standard for new DCPs.
  • Always ask for a decode report or QC log alongside the deliverable - a file that parses without error is not the same as a file that plays correctly on the target system.

The right format depends on where the content plays and what standard the destination enforces. Here is the table I use to scope a caption job before accepting it:

Delivery targetRequested formatTimecode / frame rateQC step
US broadcast (CEA-608 service)SCC (Scenarist Closed Captions)29.97 drop-framecaption-inspector decode check
US broadcast (native CEA-708 service)SMPTE-TT (ST 2052-1 XML)29.97 drop-framecaption-inspector service-layer check
Streaming (Netflix, Apple TV+, etc.)TTML (IMSC 1.1 text profile) or SRTMedia time (HH:MM:SS.mmm)Platform TTSG profile validator
Web / HTML5 video playerWebVTT (.vtt)Media time (HH:MM:SS.mmm)W3C WebVTT cue parser
DCP cinemaSMPTE ST 428-7 subtitle XMLReel edit rate (24 fps / 25 fps)DCP validator (Clipster, DCP-o-matic)

A few ground rules that apply across the whole table:

  • SCC carries only CEA-608 data: a 32-column Latin-character grid, pop-on, roll-up, or paint-on modes. It cannot carry native CEA-708 styling or any non-Latin script.
  • SMPTE-TT (ST 2052-1) is an XML format based on TTML. It is the FCC-recognised safe-harbor interchange format for CVAA-compliant online video and can encode the full CEA-708 styling set.
  • TTML/IMSC is the streaming world's preferred delivery format. IMSC 1.1 defines a text profile and an image profile; streamers almost always want the text profile.
  • WebVTT is specified by the W3C and is the only subtitle format with native CSS ::cue styling support in browsers.
  • DCP subtitle XML uses SMPTE ST 428-7 (or the older Interop schema). These files must be wrapped inside a DCP package, not delivered standalone.

US broadcast: when does SCC still matter?

SCC (Scenarist Closed Captions) is a binary format that encodes CEA-608 caption bytes as pairs of two-byte hex codes on a 29.97 drop-frame timecode grid. Despite being introduced in the early days of caption authoring, SCC is still the most common deliverable for US linear broadcast. Post-production facilities, news playout systems, and most broadcast MAM workflows expect an SCC sidecar alongside the video.

The key constraints to verify on a received SCC file:

  • Timecode base must match the video: 29.97 drop-frame (DF). A non-drop (NDF) timecode drifts roughly 3.6 seconds per hour, which is audible by any QC reviewer.
  • SCC carries CEA-608 only. Its character set is an extended Latin set (EIA-608-B Annex A/B/C). It cannot encode CJK, Arabic, or other non-Latin scripts.
  • Maximum 32 characters per row, 15 visible rows. Any text exceeding the grid wraps or is silently truncated depending on the encoder.

When a broadcast spec asks for CEA-708 positioning and styling beyond what 608 supports, the deliverable is SMPTE-TT (ST 2052-1). Open-source tools can generate this format: Subtitle Edit exports it as SMPTE-TT2052 via the SeConv command-line tool.

Streaming and VOD: which format should I request?

Streaming platforms each publish a Timed Text Style Guide (TTSG) that specifies the exact format and profile they accept. Netflix, for example, publishes requirements for over 30 languages, and each language page can specify a different character-per-line limit. Before ordering a deliverable for a streamer, download their TTSG and confirm the format requirement. What follows is what I see in practice:

  • TTML (IMSC 1.1 text profile): required or preferred by most major streamers for content requiring precise positioning and styling. The file uses XML, declares the IMSC namespace, and carries all cues inside a <body> element with media time offsets (HH:MM:SS.mmm, no drop-frame semicolon).
  • SRT (SubRip Text): widely accepted by streaming platforms as a simpler alternative where styling is not required. SRT uses sequential index numbers, media-time timestamps, and plain text dialogue lines. No styling metadata is carried.
  • Never deliver broadcast-format SCC to a streaming platform. The CEA-608 hex encoding is not interpretable by streaming ingest pipelines as a timed-text subtitle track.

A minimal TTML/IMSC cue block looks like this:

<?xml version="1.0" encoding="UTF-8"?>
<tt xml:lang="en"
    xmlns="http://www.w3.org/ns/ttml"
    xmlns:tts="http://www.w3.org/ns/ttml#styling"
    xmlns:ttm="http://www.w3.org/ns/ttml#metadata">
  <head>
    <styling>
      <style xml:id="s1" tts:fontFamily="Arial" tts:fontSize="100%"
             tts:color="white" tts:textOutline="black 3px"/>
    </styling>
    <layout>
      <region xml:id="r1" tts:origin="10% 80%" tts:extent="80% 15%"/>
    </layout>
  </head>
  <body>
    <div>
      <p begin="00:00:01.000" end="00:00:04.000" region="r1" style="s1">
        First subtitle line here.
      </p>
    </div>
  </body>
</tt>

Web video and HTML5: when is WebVTT enough?

For content that plays in an HTML5 <video> element, WebVTT is the native choice. It is specified by the W3C, is the only caption format with native browser support via the <track> element, and supports CSS ::cue styling and cue-position settings. SRT files also play in most modern browsers via the <track> element, but without cue-position or styling support.

A minimal WebVTT file with cue positioning:

WEBVTT

00:00:01.000 --> 00:00:04.000 position:50% align:center line:85%
First subtitle line.

00:00:05.000 --> 00:00:08.000 position:50% align:center line:85%
Second line here.

When to prefer TTML over WebVTT for web delivery: if the platform layers a custom player (not a raw HTML5 <video> tag), it may parse TTML directly for more precise positioning. Check the player's documentation before specifying the format.

DCP cinema: request SMPTE ST 428-7 subtitle XML

A Digital Cinema Package (DCP) carries subtitle and caption data as a sidecar XML file, referenced in the DCP's composition playlist. Two schemas exist:

  • SMPTE ST 428-7: the current standard. All new DCPs should use this schema. The namespace is http://www.smpte-ra.org/schemas/428-7/2010/DCST.
  • Interop subtitle XML: the legacy schema from the pre-SMPTE DCI era. Many older projectors still read it, but it is not recommended for new productions.

DCP subtitle delivery requirements that differ from broadcast and streaming:

  • Timing is expressed in real-time frames at the reel's edit rate (typically 24 fps for most features, or 25 fps for PAL-originated content). There is no drop-frame; the frame count is literal.
  • Fonts must be embedded inside the DCP package as a byte-identical reference. A font mismatch between the authoring system and the projector causes text rendering to fail silently.
  • DCP subtitle files must be wrapped in a package (using DCP packaging tools such as OpenDCP or DCP-o-matic). A standalone XML file cannot be loaded by a theatre playback server.

What QC should accompany every caption deliverable?

Receiving a file that opens without syntax error is not the same as receiving a file that decodes correctly on the target device. Here is the QC I ask for alongside every deliverable:

  • SCC or SMPTE-TT: a caption-inspector decode report (Comcast's open-source tool), showing the -C1.608 and -S1.708 output files. This confirms the 608 and 708 services decode correctly to readable text, not garbled bytes.
  • TTML/IMSC: a validator pass against the IMSC 1.1 schema and the platform TTSG profile. Netflix and other streamers reject files automatically on ingest if they fail their schema check.
  • WebVTT: a cue parser run confirming there are no malformed timestamps, missing cue separators, or non-UTF-8 bytes.
  • DCP XML: a DCP validator run (DCP-o-matic Verifier or the Clipster check) confirming font embedding, namespace correctness, and frame-count alignment.

I include a brief QC report with every caption deliverable I produce. For a full breakdown of what a professional QC pass covers, see the caption QC service page.

FAQ

Can I deliver SCC to Netflix or a streaming platform?

No. SCC encodes CEA-608 data as binary hex pairs and is designed for broadcast playout systems. Streaming platforms expect TTML (IMSC-profile XML) or SRT. Delivering SCC to a streaming ingest pipeline will either fail validation or produce no readable text track. Always check the platform's Timed Text Style Guide for the accepted format.

What is the difference between TTML, IMSC, and SMPTE-TT?

TTML (Timed Text Markup Language) is the base W3C XML vocabulary for timed text. IMSC 1.1 is a constrained profile of TTML defined by the W3C for internet subtitle delivery worldwide. SMPTE-TT (ST 2052-1) is a separate profile of TTML defined by SMPTE for US broadcast and CVAA-compliant online video. All three share the TTML XML syntax, but differ in which features are permitted, required, or forbidden.

Does WebVTT work for broadcast?

No. WebVTT is a web-only format intended for the HTML5 <track> element. It has no path into a broadcast playout chain, which expects SCC or SMPTE-TT for CEA-608/708 services. Delivering a WebVTT file to a broadcast station will result in the file being unusable.

When should I request SMPTE ST 428-7 versus Interop for a DCP?

Always request SMPTE ST 428-7 for new productions. Interop subtitle XML is a legacy format that predates the SMPTE standardization of DCP. Most modern projectors and servers (from 2012 onward) support SMPTE. The only reason to request Interop today is if the DCP must play on a very old projector system that has not been updated. Confirm with the exhibition venue before choosing.

What happens if I deliver the wrong format?

At best, the platform or facility rejects the file on ingest and you revise. At worst, the file passes validation but decodes incorrectly: garbled characters, wrong timing, or no text at all, discovered only during a QC screening. This is why the QC decode step is not optional. Each format has its own failure mode, and a decode report catches them before they reach the audience.

Official Sources

Dao Huy (Lucas)
Dao Huy (Lucas)
岘港大学外国语大学毕业。每份认证翻译均附签名的准确性声明(USCIS、IRCC)。

作者:Đào Huy(Lucas),越南语译员与本地化专家(英、中、法译越)。 See translation services →

报价WhatsApp