AI video with sound, 10 seconds: cheapest on Sume (Kling, Omni, H3)

Short answer

For a 10 second clip with sound, Sume's cheapest listed route is minimax-h3 at 768p for $0.75, then minimax-h3-max at 768p for $1.00, gemini-omni-flash-1.1 at 720p for $1.25 and kling-3 with audio on for $2.10. Audio is native on H3, H3 Max and Omni, and is a flag on Kling 3 that raises the rate from $0.14 to $0.21 a second. Rates are provider list times 1.25.

The table

Per-second Sume rates from the pricing tables, multiplied by ten. H3 and H3 Max use 768p, not 720p, as their native size.

What "with sound" means on each

H3 produces native stereo audio and always has it on; sending generate_audio: false is rejected. Omni has native synced audio. Kling 3 has an audio switch, so the same clip is $1.40 with it off and $2.10 with it on. If you need silence, Kling 3 is the one where you save money by switching off; on H3 you cannot.

Choosing

Price is only one factor. Omni accepts 16:9 and 9:16 and 3 to 10 seconds, so it fits a 10 second cap exactly. H3 accepts 5 to 15 seconds and more aspect ratios, which helps for 21:9 or longer clips. Kling 3 costs more per second but is the model to test if motion style is what you are paying for. Run the same prompt on each, then compare, since a cheaper clip you reject is not cheap.

Submit one

Swap the model id and resolution to compare:

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"minimax-h3","prompt":"A busker plays violin in a train station","duration":10,"resolution":"768p"}'

Checking the audio itself

Generated audio is part of what you pay for, so listen before you compare prices. Play each clip with headphones for ambience, speech clarity and any glitches at the cuts. If you only need a music bed, a silent Kling 3 clip at $1.40 plus separate audio may serve better than a clip whose audio you will replace anyway. Decide what the sound is for, then compare.

Caveats

Wan 3.0 and Seedance also carry audio options, but their audio pricing was not part of this comparison. The usage.cost on a finished job is the billed amount. Reference images beyond the first five on H3 add a surcharge.