30 seconds of footage: Omni, H3 Max with audio, or Wan plus a bed

For 30 seconds of footage with sound, the cheapest route on Sume list prices is MiniMax H3 Max at 768p: two 15-second clips with native stereo audio, $3.00 of footage plus a $0.10 render for $3.10. Gemini Omni Flash 1.1 at 720p is $3.75 for 30 seconds, plus the render, for $3.85. One Wan 3.0 720p clip of 30 seconds plus a Music bed is $3.75 + $0.125 + $0.10 = $3.975.

The three routes differ in where the sound comes from. The Sume docs say Omni's native audio is always on and that H3 Max renders with native stereo audio. A Wan clip has no bed unless you add one. This post gives only the prices and the structure; it does not rank how the footage looks or sounds.

The arithmetic

Per-second rates are Sume list prices read from the public catalog on 2026-10-08: Omni 720p $0.125, H3 Max 768p $0.10, Wan 3.0 720p $0.125. Omni accepts 3 to 10 seconds, H3 Max 5 to 15 seconds and Wan 3.0 2 to 30 seconds, so the clip counts follow the maximum length. A Timeline 1.0 render is $0.10 per started output minute and is needed whenever clips are joined or a bed is mixed in.

What the bed adds, and what it does not

The Wan route is the only one that needs a bed to have any sound. The other two already have audio, so a bed there is a choice. If you add one, you also take on a mixing decision: the Timeline page documents the soundtrack bed with gain_db, loop, fade_out_seconds and duck_db, but it describes the duck as running under a spine, not under clip audio. Read the Timeline page on how your clips' own audio is treated before you stack a bed on top of native audio.

If you want to look at the native audio on its own, audio detach takes the audio of one Sume-hosted video as a wav or mp3 for $0.01 per job. That is the way to listen for unwanted music or speech before you add a bed to a clip that already has sound.

Choosing

Pick H3 Max when native stereo is wanted and the cheapest footage matters. Pick Omni when the 3 to 10 second shot length and the 16:9 or 9:16 ratio fit the cut. Pick Wan when you need a single 30-second take, because one clip avoids the cut points that the other two routes need. In the Wan route the bed is 3.1% of the $3.975 total ($0.125 / $3.975), so the sound is not where the money goes.

A final check before you commit. Each model has a minimum billed length and its own valid range, and the Video Router catalog (GET /v1/video-router/models) lists the live envelope. Confirm the clip lengths there before you queue a batch.