Embed CEA-608 Captions in H.264 SEI with FFmpeg

This guide demonstrates how to construct an FFmpeg command to stream live H.264 video with embedded CEA-608 closed captions within the SEI (Supplemental Enhancement Information) NAL units. You will learn the required FFmpeg parameters, how to handle different caption sources, and the exact command-line syntax needed to deliver broadcast-compliant live streams to platforms like YouTube, AWS Elemental MediaLive, or Facebook Live.

The Core Parameter: a53cc

To embed CEA-608 captions into the H.264 video stream as SEI user data, you must use the libx264 encoder with the a53cc option enabled. The A/53 specification defines how closed captions (specifically CEA-608 and CEA-708) are packaged inside the MPEG-2 and H.264 video bitstreams.

By setting -a53cc 1 (or -x264-params a53cc=1), FFmpeg looks for caption side data in the input stream and injects it directly into the output H.264 SEI NAL units.

Scenario 1: Streaming with an External SCC Caption File

If you have a live video input and a synchronized SCC (Scenarist Closed Caption) file containing your CEA-608 data, use the following FFmpeg command to merge them and stream via RTMP:

ffmpeg -re -i input_video.mp4 -i captions.scc \
  -map 0:v:0 -map 0:a:0 -map 1:s:0 \
  -c:v libx264 -pix_fmt yuv420p -profile:v main -g 60 \
  -a53cc 1 \
  -c:a aac -b:a 128k \
  -f flv rtmp://live.server.com/app/stream_key

Command Breakdown:

• -re: Forces FFmpeg to read the input at native frame rate, which is required for live streaming.
• -i input_video.mp4 -i captions.scc: Imports both the video/audio source and the external caption file.
• -map 0:v:0 -map 0:a:0 -map 1:s:0: Maps the first video stream, the first audio stream, and the subtitle stream from the SCC file into the output.
• -c:v libx264: Encodes the video to H.264.
• -a53cc 1: Instructs the encoder to embed the mapped subtitle data as A/53 SEI closed captions.
• -c:a aac: Encodes the audio to AAC, the standard for RTMP streaming.
• -f flv: Formats the output container as FLV for RTMP ingest.

Scenario 2: Streaming from a Source with Pre-Existing Captions

If your input source already contains caption side data (such as an SDI capture card, an MPEG-TS feed, or an MXF file with EIA-608 data), you do not need an external caption file. FFmpeg will automatically read the side data from the input and write it to the output:

ffmpeg -re -i input_with_captions.ts \
  -c:v libx264 -pix_fmt yuv420p -preset medium \
  -x264-params "a53cc=1" \
  -c:a aac -b:a 128k \
  -f flv rtmp://live.server.com/app/stream_key

Command Breakdown:

• -x264-params "a53cc=1": An alternative syntax to pass the a53cc flag directly to the libx264 encoder library, ensuring compliance across different FFmpeg builds.

Verifying the Captions in the Output

To verify that your output stream successfully contains the CEA-608 SEI captions, you can run a quick diagnostic command on a saved segment of your stream using the ffprobe tool:

ffprobe -show_entries frame=side_data_list -select_streams v -v quiet output.mp4

Look for "side_data_type": "Closed captions" in the JSON output. If this block is present for your video frames, the CEA-608 data has been correctly embedded into the H.264 SEI NAL units.