Reel voiceover too long for 60 seconds? Speed setting and re-render

If your voiceover runs 68 seconds and the Reel should be 60, ask for a read about 13 percent faster: generation_config.speed on Sume's TTS 1.0 accepts 0.6 to 1.5. The multiplier is the current length divided by the target length, and a regeneration at $0.0475 per 1,000 characters is cheap enough to try twice.

Instagram's Reels guide (read 2026-10-03) calls 15 to 60 seconds the sweet spot for tutorials, day-in-my-life clips and most creative formats, which is where this fit problem usually arises.

The arithmetic

Speed is a multiplier on the pace, so new duration is about old duration divided by speed. This is a planning estimate, because Sume's docs do not publish a guaranteed relation between speed and length; read the real duration_seconds from the new job.

Compute it and re-submit

The helper below returns the speed to try, or None when the gap is beyond the accepted range, which is the signal to cut the script instead.

def speed_for(current_s, target_s, lo=0.6, hi=1.5):
    s = round(current_s / target_s, 2)
    return s if lo <= s <= hi else None

for cur in (62, 68, 75, 90, 100):
    print(cur, speed_for(cur, 60))

When to rewrite instead

Above roughly 1.25 a read starts to sound hurried to most listeners, so past that point trimming a sentence usually serves the Reel better; that is a listening judgment, not a Sume limit. A script change costs one new TTS job. If you have already cut the voice into parts, regenerate only the part that overran and join again with timeline audio concat, a flat $0.01.

What Sume does not do

Sume does not time-stretch a finished file to fit, and it does not shorten a script for you. Speed is a generation setting: change it and you get a new job and a new duration_seconds. Feed that number into Timeline 1.0 as audio.duration_seconds so the render, and the $0.10 per output minute, match the voice you have.