720p to 1080p: Veo 3.1 Fast adds 20%; H3 Max and Wan double on Sume

Whether to render at 1080p instead of 720p is partly a price question, and the premium differs a lot by model. On Google's Gemini API pricing page (read 2026-10-03), Veo 3.1 Fast is $0.10 per second at 720p and $0.12 at 1080p, a 20 percent premium. Veo 3.1 Standard is listed at $0.40 at both resolutions, so the premium is zero.

The premium on Sume

Sume's video rates (provider list times 1.25, see the Video API docs) rise more steeply. MiniMax H3 Max goes from $0.10 per second at 768p to $0.20 at 1080p. Wan 3.0 goes from $0.125 at 720p to $0.25 at 1080p. Both are exactly a 100 percent premium. For H3 Max, the docs describe 1080p as a latent refinement from native 768p, so you are paying for the refinement step.

What to do with that

If you buy on Sume, draft at the lower resolution and render only the clips you keep at 1080p. For 20 candidate clips of 10 seconds, drafting on H3 Max at 768p is $20, and re-rendering the 5 you keep at 1080p is another $10, for $30. Rendering all 20 at 1080p straight away is $40.

On Google's list the same trick saves almost nothing, since the Fast premium is only $0.20 per 10 seconds. The cheaper habit there is to pick a tier, not a resolution. Note that Sume's H3 Max 1080p is a refinement of the 768p result, not a second independent take, so the draft and the final look related.

Check it yourself

The snippet computes the draft-then-final cost for any counts.

draft_rate, final_rate = 0.10, 0.20  # H3 Max 768p and 1080p on Sume, per second
seconds, candidates, kept = 10, 20, 5
draft = candidates * seconds * draft_rate
final = kept * seconds * final_rate
print(f'draft {draft:.2f} + final {final:.2f} = {draft + final:.2f}')
print(f'all at 1080p: {candidates * seconds * final_rate:.2f}')