Walmart silent video: the [NO SPEECH] caption and Kling audio off

A Walmart Marketplace video with no music, audio or sound should start with a "[NO SPEECH]" caption that can come off after about 10 seconds. On Sume, a 10-second Kling Video v3 Pro clip with audio off costs $1.40, and a cue-based caption job adds $0.20, so one compliant silent clip costs $1.60 before any retries.

What Walmart's PDF says

Page 2 of Walmart's Rich media: Technical requirements, listed as last updated on Sep 17, 2026 (read 2026-10-08), gives the best practice for a video that has no music, audio or sound: insert a caption that says "[NO SPEECH]" when the video begins, and remove it after about 10 seconds. The same page gives the music-only variant, "[MUSIC PLAYING]".

The PDF does not say the video must be silent or must have sound. It only says what to write when it has none.

Making the silent clip on Sume

The video-router catalog lists Kling Video v3 Pro as kling-3 with 720p and 1080p, 16:9, 9:16 and 1:1, and durations of 4 to 15 seconds. It has an audio toggle. List price is $0.112 a second with audio off and $0.168 with audio on, and Sume sells every listed row at list times 1.25, rounded up to the cent per job.

Audio off therefore bills $0.14 a second, which is $1.40 for 10 seconds. Grok Imagine Video 1.5 is cheaper at $0.0125 a second, but it is image-to-video only, has no audio, and takes an image_url or first_frame_url.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: walmart-no-speech-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/silent.mp4",
    "cues": [
      { "text": "[NO SPEECH]", "start": 0, "end": 10 }
    ]
  }'

Checks before you upload

Cues skip speech recognition, which is why they are the path for silent clips: the caption docs say a silent clip run through speech-to-text fails with caption_no_speech. Keep cue timings inside the clip length.

Walmart lists the other limits on page 1: the file must be .mp4, and the 100MB limit is typically but not always accepted. A 10-second clip is far below that, but check the final file size.

If your catalog mixes silent and music clips, sort the batch before you caption it. A silent clip gets [NO SPEECH], a music clip gets [MUSIC PLAYING], and a clip with speech gets ordinary captions. Sorting is a spreadsheet column, and the caption call takes the text from your row.

• Audio off on Kling 3 is a render setting. It does not strip sound from a clip you already have.
• The PDF lists no minimum clip length, so do not assume one.
• Grok Imagine Video 1.5 at 480p and 720p only. Check the resolution you need against the catalog.