Shorts 2x playback speed: captions flash by, so set phrasing max_words

Short answer

At 2x playback a caption phrase that was readable for 1.2 seconds is on screen for 0.6. Burn captions in short phrases: in Sume's video captions API set design.phrasing.max_words to a low number such as 3, which works with the slam style. Punch and tiktok-green do not support design overrides.

What YouTube announced

YouTube's blog post of June 25, 2026 on Shorts experience updates lists a heart that replaces the thumbs-up, a retired dislike button, Clear Screen that hides icons and text, 2x playback speed, a Shorts timer and Not Interested reasons (read 2026-10-05). The post does not describe how many viewers use 2x, so treat it as a possibility, not an average. A burned-in caption cannot be sped up or slowed down by the viewer, so it has to be readable at the fastest speed you care about.

The reading-time arithmetic

On-screen time at 2x is half the time at 1x. The phrase times and the reading rate below are assumptions for the arithmetic, not data about viewers.

A request with short phrases

The 3 words per second figure is an assumption, not a measured reading speed. With it, a 3-word phrase needs about 1 second to read, which is exactly what a 2-second phrase gives you at 2x. Limit the phrase to three words with design.phrasing.max_words and each card is short enough to survive the faster speed. The request body below changes nothing else about the slam style. Phrase length is only one lever; a clip with long pauses between sentences also gives each card more time on screen than a clip with continuous speech.

import json, os, urllib.request

key = os.environ.get("SUME_API_KEY", "")
if not key:
    raise SystemExit("set SUME_API_KEY")
body = {
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "style": "slam",
    "design": {"phrasing": {"max_words": 3}},
}
req = urllib.request.Request(
    "https://api.sume.com/v1/video-captions",
    data=json.dumps(body).encode(),
    headers={"Authorization": "Bearer " + key,
             "Content-Type": "application/json",
             "Idempotency-Key": "caption-fast-001"},
)
print(json.load(urllib.request.urlopen(req)))

What else the same update affects

Clear Screen hides icons and text, so check that your burned-in captions are part of the picture and do not rely on an overlay the viewer can hide. If the caption line collides with on-screen controls on your own test device, move it with design.placement.anchor_ratio, the center of the line as a fraction of frame height.

A caption job is $0.20 for a video up to 60 seconds. Restyling with source_caption_id re-burns the same words with a new look at the same price, so you can try max_words 2 against 3 without a second transcription.

A short test plan: burn the same clip with max_words 3 and max_words 2, play both at 2x on a phone, and keep the one you can read without pausing. Do the same with a clip that has fast speech, since that is where phrase cards turn over fastest. If neither is readable, the fix is upstream: slow the speech or cut the script, because a caption cannot hold a card on screen longer than the words are spoken.