Voiceover or video: which is the bigger cost in a 15-second ad?

The clip is the bigger cost by two orders of magnitude. A 250-character voiceover is $0.0119 on Sume TTS ($0.0475 per 1,000 characters), while a 15-second Wan 3.0 clip at 720p is $1.88 (fal list $0.10 per second, read 2026-10-09, x 1.25).

What are the line items?

The 250-character script is an assumption for roughly a 15-second read; change it to match yours. Each row uses Sume's list x 1.25 price.

What share is the voice?

At 480p the voiceover is 1.3% of the pair, at 720p 0.6% and at 1080p 0.3%. Retaking the voice five times costs less than one more 480p clip.

So what should I retake?

• Re-record the voice freely: ten takes of a 250-character line is $0.119.
• Lock the script before generating video, because a changed line can change the length the clip must cover.
• Spend the retake budget on the picture: three extra 720p takes are $5.625.

Does anything else belong in the sum?

Yes, a final render. Timeline 1.0 is priced at $0.10 per output minute (rounded up) in the Timeline 1.0 docs, so a 15-second ad adds $0.10. That is still smaller than the clip.

What about a 60-second video ad?

Scale it up: a 60-second ad has about 1,000 characters of script, which is $0.0475 on Sume TTS. The video is four 15-second Wan 3.0 clips: $3.75 at 480p, $7.50 at 720p or $15.00 at 1080p. The voice is about 1% of the sum at 480p and less at the higher tiers, so the cost conversation is always about the picture.

If you also keep five script variants, voice them all: five 250-character reads are $0.0594 on Sume TTS, less than the cost of one second of 720p Wan 3.0 video ($0.125). That makes script testing nearly free next to picture testing. Run the cheap loop first: choose the script on audio alone, then spend the video budget on the chosen line.