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。
本地化笔记🎬 字幕与隐藏字幕

IMSC and TTML Subtitles: When Streaming Platforms Require Them

Netflix, Disney+ and Amazon require TTML/IMSC for subtitle delivery. Here is what TTML carries that SRT cannot, and how to convert files with one SeConv command.

Dao Huy (Lucas)6 min read7 节
IMSC and TTML Subtitles: When Streaming Platforms Require Them

💡 SRT works for YouTube and web delivery, but Netflix, Amazon Prime, Disney+ and Apple TV+ delivery pipelines typically require TTML - an XML subtitle format also called IMSC or DFXP. TTML carries named positioning regions and styling metadata that SRT cannot express. Convert any SRT to IMSC 1.1 in one command: seconv in.srt TimedTextIMSC1.1 --output-folder out

Key takeaways

  • SRT has no region or positioning data; TTML carries named regions (such as region.bottomCenter) with percentage coordinates - this is why professional delivery specs require TTML.
  • Netflix standard subtitle format: TTML/DFXP (SeConv format name: NetflixTimedText, extension .dfxp). Disney+ and IMSC-requiring platforms: TimedTextIMSC1.1 (extension .xml).
  • YouTube, Vimeo and self-hosted web video accept SRT or WebVTT - TTML is not required there.
  • Use percentage units, not pixels: Netflix and most streaming TTML specs require tts:fontSize="100%" and percentage-based region coordinates.
  • The ttp:profile URI on the <tt> root element declares which IMSC profile the file conforms to; SeConv sets it correctly when you pick the right format name.

What is TTML and where does IMSC fit in?

TTML (Timed Text Markup Language) is a W3C XML-based subtitle standard. The W3C published TTML2 as the current full recommendation, but for OTT video delivery the relevant subset is IMSC (Internet Media Subtitles and Captions) - a constrained profile of TTML designed so that all compliant players implement the same feature set reliably. IMSC is what I work with for every professional subtitle delivery.

DFXP (Distribution Format Exchange Profile) was the original name for TTML 1.0's distribution variant. The two terms appear interchangeably in platform delivery specs. When Amazon Prime or an older broadcast spec says "DFXP", they mean TTML.

IMSC defines two profiles: the text profile for Unicode text subtitles, and the image profile for bitmap subtitle deliveries. Text profile is the standard choice for virtually all dialogue subtitles. The profile is declared with the ttp:profile attribute on the <tt> root element.

What can SRT not carry?

SRT is a minimal format: a sequence number, a timing line, and plain text. It carries no region, no positioning, and no structural styling beyond basic tags like <i>. This is fine for YouTube or Vimeo, where the player handles layout. It becomes a problem for professional delivery where the platform's ingest system needs to verify:

  • Region and positioning: should this cue render bottom-center or top-right? SRT cannot specify.
  • Forced narratives: is this a forced subtitle (shown even when the subtitle track is off)? SRT has no flag for this.
  • Language and profile declaration: the xml:lang attribute on the <tt> root and ttp:profile make a TTML file self-describing.
  • Styling metadata: TTML regions carry tts:textAlign, tts:displayAlign, and font metadata the platform validates at ingest.

Which streaming platform wants which format?

This is the decision table I work from when scoping a delivery. Always confirm against your distributor's current delivery specifications before final submission - platform requirements do change.

Delivery targetFormatExtensionSeConv format name
Netflix (non-Japanese content)TTML / DFXP.dfxpNetflixTimedText
Netflix (Japanese content)IMSC 1.1.xmlNetflixIMSC1.1Japanese
Amazon Prime VideoDFXP/TTML or SRT.dfxp / .srtNetflixTimedText or SubRip
Disney+IMSC 1.1.xmlTimedTextIMSC1.1
Apple TV+ / HLS deliveryIMSC 1.1 in fMP4.xml (then packaged)TimedTextIMSC1.1 (then package)
YouTube / Vimeo / webSRT or WebVTT.srt / .vttSubRip or WebVTT
US broadcast (sidecar)SCC (608) or SMPTE-TT (708).scc / .xmlScenaristClosedCaptions or SMPTE-TT2052

Note: US broadcast closed captioning uses CEA-608 (SCC) or CEA-708 (SMPTE-TT) - completely separate standards from TTML/IMSC. SCC carries English captions only and cannot encode CJK characters. TTML/IMSC is the streaming XML track format.

How do I convert an SRT to IMSC 1.1 with SeConv?

I use SeConv (the Subtitle Edit batch converter) for all TTML conversions. The format name for IMSC 1.1 output is TimedTextIMSC1.1 (outputs .xml):

seconv in.srt TimedTextIMSC1.1 --output-folder out --overwrite

For a Netflix TTML/DFXP delivery, use NetflixTimedText (outputs .dfxp):

seconv subtitles.srt NetflixTimedText --output-folder delivery --overwrite

Both accept a glob pattern to convert a whole folder in one run:

seconv "subtitles\*.srt" TimedTextIMSC1.1 --output-folder delivery --overwrite

The output from TimedTextIMSC1.1 is a well-formed TTML file. Here is an annotated example of the structure SeConv generates (verified by running it locally against a test SRT):

<?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:ttp="http://www.w3.org/ns/ttml#parameter"
    ttp:profile="http://www.w3.org/ns/ttml/profile/imsc1/text"
    ttp:timeBase="media">
  <head>
    <styling>
      <style xml:id="style.center"
             tts:color="#ffffff"
             tts:fontSize="100%"
             tts:fontFamily="default"
             tts:textAlign="center" />
    </styling>
    <layout>
      <region xml:id="region.bottomCenter"
              tts:origin="17.583% 73.414%"
              tts:extent="64.844% 16.667%"
              tts:displayAlign="center"
              tts:textAlign="center" />
    </layout>
  </head>
  <body>
    <div xml:id="d0" style="style.center">
      <p begin="00:00:01.000" end="00:00:04.500"
         region="region.bottomCenter">First subtitle line.</p>
      <p begin="00:00:05.200" end="00:00:08.100"
         region="region.bottomCenter">Second subtitle line.</p>
    </div>
  </body>
</tt>

Understanding the key attributes in this output helps you spot validation problems:

  • ttp:profile="http://www.w3.org/ns/ttml/profile/imsc1/text": declares this file conforms to the IMSC1 text profile. The player uses this URI to know which features to expect. SeConv sets it when you choose TimedTextIMSC1.1.
  • ttp:timeBase="media": timing is relative to the media clock - standard for OTT delivery. Every streaming ingest system expects this.
  • tts:fontSize="100%": percentage-relative font size, not pixels. Netflix's TTML spec mandates percentages; a pixel value here triggers a validation failure at ingest.
  • tts:origin="17.583% 73.414%": the top-left corner of the region as a percentage of the root container. This positions the region roughly 73% down from the top - the standard bottom-center placement.
  • tts:extent="64.844% 16.667%": the width and height of the region in percentages. The region spans about 65% of the frame width, centered horizontally.
  • begin="00:00:01.000" end="00:00:04.500": timing on each <p> in HH:MM:SS.mmm format, relative to media start. Both begin and end must be present on every subtitle event.

What can I do myself, and when does a packaging pipeline help?

Converting SRT to TTML with SeConv is a one-command step I do on every professional delivery - no programming needed. For standard dialogue subtitles at bottom-center with correct timing, the SeConv defaults produce a submission-ready file. This covers Netflix and Amazon deliveries in my experience with English subtitle tracks.

Where the work gets more specialized:

  • Apple TV+ / HLS: requires IMSC 1.1 packaged inside ISOBMFF fMP4 fragments as a subtitle segment track. The packaging step is beyond SeConv and needs Bento4, an ffmpeg HLS pipeline with -c:s stpp, or a commercial packaging service.
  • Non-English subtitle translation: the TTML structure is correct, but the words still need a human translator who handles reading speed, line breaks, and linguistic accuracy. That is where my caption QC review catches timing and language issues before final delivery.
  • Multi-track deliveries: separate forced-narrative, SDH, and dialogue-only files in TTML, each validated against the platform's current TTSG, require careful workflow design to avoid timing drift between tracks.

FAQ

Is TTML the same as IMSC?

IMSC is a constrained subset (a profile) of TTML. All IMSC files are valid TTML, but not all TTML files conform to IMSC. IMSC removes ambiguous or rarely-implemented features so that all compliant players behave consistently. When a streaming platform says "TTML" in their delivery spec, they typically mean TTML conforming to the IMSC text profile.

Can I open and edit a TTML/IMSC file in Subtitle Edit?

Yes. Subtitle Edit reads TTML, DFXP and IMSC files directly. Open the file, edit cue text and timing in the standard interface, then export back using File > Save As or the SeConv batch command. The XML structure is handled transparently and you see plain subtitle events.

Netflix rejected my TTML file - what do I check first?

Three common rejection causes in my experience: (1) pixel units in region coordinates or font size - switch all values to percentages; (2) a missing or mismatched xml:lang on the <tt> element vs the declared language in the delivery package; (3) overlapping cue timing or zero-gap between consecutive events. Run the file through Subtitle Edit's "Fix common errors", re-export with NetflixTimedText, and re-submit.

What is DFXP - is it different from TTML?

DFXP (Distribution Format Exchange Profile) is the original name for TTML 1.0's distribution variant. The two terms are used interchangeably. When Amazon Prime or older platform specs say "DFXP", they mean TTML. The .dfxp extension and the .ttml extension point to the same format family.

Do I need TTML for YouTube subtitles?

No. YouTube accepts SRT and WebVTT for subtitle uploads, and those formats work well for web delivery. TTML/IMSC becomes a requirement specifically for professional delivery to subscription streaming platforms (Netflix, Disney+, Amazon) and for Apple's HLS packaging pipeline. Use SRT or WebVTT for YouTube.

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