A 60-second explainer: voice, six stills, six clips on Sume vs list

A 60-second explainer built from AI parts has three cost lines: video, stills and voice. Here is one concrete build. Six 10-second clips for 60 seconds of video; six Nano Banana 2 stills at 1K as first frames; and an 800-character voiceover, which is about a minute at Cartesia's stated 750 to 800 characters per minute.

The comparison is Sume against the same build bought at list from Google (Gemini API pricing, Veo 3.1 Fast and Nano Banana 2) and ElevenLabs (API pricing, v4 launch price), read 2026-10-03.

The three lines

On Sume: H3 Max at 768p is $0.10 per second, so six 10-second clips are $6.00; six Nano Banana 2 stills at 1K are $0.60 at $0.10; and 800 characters at $0.0475 per 1,000 is $0.038. At list: Veo 3.1 Fast at 720p is $0.10 per second, so $6.00; Nano Banana 2 at 1K standard is $0.40; and ElevenLabs v4 at its launch price of $0.022 is $0.0176.

What drives the total

Video is about 90 percent of the bill either way. The stills and voice are rounding error next to the clips, so the two columns differ by only the small gap in the stills ($0.60 against $0.40). The ElevenLabs v4 price in the table is a launch price that ends October 12; at its $0.08 list the voice line would be $0.064, still under seven cents.

That is the useful insight: when you compare vendors for a build like this, compare the clip rate, and treat stills and voice as free. Everything that changes the total is in the video line: tier, resolution, and the number of takes.

Retakes

A real project does not render each clip once. If each clip needs two takes, the video line doubles to $12.00 on both sides, and the project is $12.64. Draft at lower resolution where the model allows it, and re-render only what you keep.

The snippet prices a build and applies a retake factor.

clips, seconds, clip_rate = 6, 10, 0.10
stills, still_rate = 6, 0.10
voice_chars, voice_rate = 800, 0.0475
retakes = 1.0
video = clips * seconds * clip_rate * retakes
total = video + stills * still_rate + voice_chars / 1000 * voice_rate
print(f'total: ${total:.2f}')