Silent menu-special clip: authored captions on 12 seconds for $0.22

To caption a silent menu-special clip, send the clip with the cues field: you write the text and timing, and Sume burns them in. Captions cost $0.20 for a clip up to 60 seconds, and a video trim to 12 seconds first costs $0.02, so the pair is $0.22.

Without cues, Sume transcribes the clip's speech, so a silent file returns a caption_no_speech error (read 2026-10-06). Cues are the way around it.

The two steps

First import the source with a media import so the file lives on media.sume.com. Both tools take those URLs, and each request carries an Idempotency-Key.

Trim the clip to the beat you want, then send the trimmed result to captions with cues. Keep each cue short: a dish name, a price, a day. A viewer reads a few words in the second or two a cue stays up.

Why authored cues

Menu specials are mostly food footage with no voice. A transcription step would find nothing, and writing the text yourself guarantees the dish name and price are spelled right, which a speech model cannot promise for a menu word.

If you later want the same cues in a different style, the docs describe restyling from a previous caption job without transcribing again, and a design object can override the look.

A worked example

Say the clip shows a bowl of ramen for 12 seconds. You might write three cues: the dish name for the first four seconds, the price for the next four, and the days it is available for the last four. Each cue is a few words, readable on a phone without sound.

Check the placement. Captions are drawn on the frame, so a cue that sits over the food hides it. The captions design options let you change position and style, and the docs describe a restyle that reuses an earlier caption job instead of starting over. Try a first render and look at it on a phone before you caption the other clips.

Tradeoffs

Captions are burned into the pixels, so a price change means a new render. Keep prices in the post text if they change weekly, and use burned-in text only for the dish name and the day.