What a 30-second AI product ad costs end to end on Sume: $3.17

Three 5-second clips, a 450-character voiceover, a Timeline render and captions come to about $3.17 on Sume before the agent fee. Costs by step, and the order.

4 min readSume
All posts

A 30-second product ad built from three 5-second seedance-2-mini clips at 720p, a 450-character voiceover, one Timeline render and captions costs about $3.17 on Sume before the 5.5% agent fee that applies by default. The clips are $2.85 of that; the voiceover is about $0.02, the render is $0.10 and the captions are $0.20.

Every figure comes from a Sume source read 2026-10-01: the clip estimates from the pricing code behind the video docs, the render rate from the Timeline docs, the caption price from the captions docs and the voiceover rate from the rate card. The script length is an assumption.

What does each step cost?

The 450-character script is a typical 30-second voiceover; recompute with your own length at $0.0475 per 1,000 characters.

30-second product ad, Sume rates and estimates, read 2026-10-01.
StepSume itemCost
Three 5-second shots, 720pseedance-2-mini, 3 x $0.95$2.85
Voiceover, 450 charactersText to speech, 450 x $0.0475 per 1,000$0.02
Join clips and voiceoverTimeline, 1 output minute$0.10
Burned-in captionsStandalone caption job, up to 60 s$0.20
TotalBefore the agent feeabout $3.17

In what order do I run the steps?

Generate the clips and the voiceover first; both come back as Sume-hosted files. Send them to POST /v1/timeline-1.0/render with the voiceover as audio.url and its length as audio.duration_seconds. When the render completes, send its video_url to POST /v1/video-captions. Each step is its own job with its own Idempotency-Key.

What makes the price change?

Clip resolution and model dominate: the same three shots at 480p are $1.32, and seedance-2 at 720p is $5.67. Script length moves the voiceover by about $0.05 per 1,000 characters. Rerolled clips are billed when they complete, so budget one retry per shot while you are tuning prompts.

How do I confirm the total before spending?

POST /v1/timeline-1.0/plan returns the render estimate unbilled. For the rest, read GET /v1/catalog and compare against GET /v1/usage after the first ad. See how much an AI video costs for per-model figures.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume