Same vendor, two Seedance 2.5 prices: $0.0738/s vs $0.231 at 720p

Two pages from the same vendor, Higgsfield, show different Seedance 2.5 prices. The API blog lists Seedance 2.5 at $0.0738 per second, and the video API comparison, described there as verified in September 2026, lists it at $0.231 at 720p. The second figure is about 3.1 times the first, and neither page, as read, explains the gap.

What each page lists

The comparison page is vendor-authored, so treat rival prices on it as unverified. One row can be checked against a primary source: Google's pricing page lists Veo 3.1 Standard at $0.40 per second, and the comparison lists the same figure.

Why a gap like this appears

Plausible reasons for a gap include different units, resolutions, audio settings, or a plan-dependent rate, but the pages do not say. Do not pick the lower number for a budget on a guess. Ask the vendor for the rate that applies to your exact request, or run one paid request and read what you were charged.

Normalize before comparing

Units differ across vendors too. The Sume docs show an example catalog row priced per 1,000 video tokens for seedance-2, and the video docs say the billable amount is provider list times 1.25, reported as usage.cost on the poll response. Normalize every quote to cost per finished second before comparing, as in the stored post on normalizing per second and per token.

What the gap costs at scale

Take a 30 second clip. At $0.0738 per second it costs about $2.21. At $0.231 per second it costs about $6.93. For 100 clips that is $221 against $693, a difference of $472 from a number that no page explained.

The arithmetic is the reason to settle the question with one paid request before you commit a budget. A gap of this size would change a vendor choice, and it can sit unnoticed in a spreadsheet because each figure looks reasonable on its own page.

• 30 seconds at $0.0738: about $2.21.
• 30 seconds at $0.231: about $6.93.
• 100 clips: $221 against $693.
• Settle it with one paid pilot request.

A routine for any price table

A short routine keeps comparisons honest. Do it once per vendor, and date the result.

• Find a row you can check against the model owner's own page.
• Write down the resolution, duration, and audio setting for every number.
• Convert tokens, credits, and per-clip prices to dollars per second.
• Run one real request and compare the invoice to the table.