Cheap AI Video Drafts Before the Final Render: 360p, 480p, Grok Rates

Most of the money in AI video goes to takes you throw away. A draft tier fixes that if it is cheap enough to run five times, and if you know what a draft can and cannot tell you. These are the lowest per-second rates in Sume's catalog on 2026-10-03.

Draft rates

Sume bills provider list times 1.25. For a 5-second clip:

Draft to final

Take one shot: draft five variants of a 5-second clip on Omni at 360p for about $0.94, pick one, then render it at 1080p at $0.1875 per second for $0.94. The whole shot costs under $2. Rendering five finals directly would cost about $4.69. Google's own Gemini API page lists Omni as token-priced at $17.50 per million video output tokens, roughly $0.10 per second at 720p, which is the provider rate behind Sume's tiers.

The saving is the point; the catch is fidelity.

What a draft can and cannot tell you

A draft shows framing, motion direction and whether the prompt is understood. It does not promise the final's detail. Sume exposes no seed on any video model, so the final is a new generation from the same prompt, not an upscale of the draft. Treat the draft as a casting call: it picks the idea, and the final may differ in the details.

Resolution tiers are model-specific: Omni goes 360p to 4K, H3 Max 480p to 1080p, Wan 480p to 1080p. Check GET /v1/videos/models and pass resolution explicitly, since Auto defaults to 720p and 8 seconds.