Fit a 330-character script into a 20-second slot: edit, speed, extend

A 330-character script is about 22 seconds at an assumed 15 characters per second, two seconds too long for a 20-second slot. On Sume there are three fixes, in order of cost: cut about 30 characters from the text, set a faster speed, or make the slot longer. Each TTS take of 330 characters costs 330 x $0.0475 / 1,000 = $0.0157, so even three takes are under 5 cents.

Fix 1: edit the text

The cheapest fix is a shorter script. Cutting from 330 to 300 characters removes about 2 seconds at the assumed pace, and a shorter read nearly always sounds better in an ad. It also costs less: 300 characters is $0.01425. Do this first, then measure the file you get.

Fix 2: set the speed

The TTS request accepts a top-level speed of slow, normal or fast, and generation_config.speed from 0.6 to 1.5. If speed scales duration in inverse proportion (an assumption; check it on your voice), 22 seconds at 1.1 is 20 seconds. Send only one of the two speed controls, and read the value back from the finished job, which records speed and generation_config. A fast read at 1.5 is a different style of read, not just a shorter file.

Fix 3: extend the slot

In a Timeline render the spine decides the output length: audio.duration_seconds is 1 to 1,800. Slot lengths come from video[].duration, at least 0.2 seconds, and coverage may stop at most 0.5 seconds before the end of the spine. If the platform allows a 22-second video, change the slot and keep the read. The render price does not move: both lengths reserve 1 minute, which is $0.10.

Which fix suits which video

Pick by what the viewer sees. For a talking-head ad, a fast read hurts most because lips and pace are on screen; trim the words instead. For a product montage with a voice over stills, a faster read is hardly noticed and the slot can stay. For a hard platform limit, take the text edit, since it is the only fix that cannot make the clip longer.

Whatever you pick, keep the first take. If the trimmed one sounds worse, you have already paid $0.0157 for the original, and the second take is another $0.01425, so a trial of both costs about 3 cents. That is cheaper than arguing over which is better.

Measure before you decide

Four checks stop you from fixing the wrong thing.

• Characters per second is an assumption. Measure your first take; the real number may be 13 or 17.
• Spaces and punctuation count toward the Sume price, so editing punctuation barely changes the bill, but it changes the pauses.
• If the 2-second gap is silence at the end, you may not need any fix; trim the spine with source_in, or end the video on the last word.
• If the clip is a platform ad with a hard cap, check that platform's own current rule before you choose which fix to use.