Sound surcharge per second: Kling 3 and three Veo 3.1 rows on Sume

Four Sume video rows bill sound as a separate rate: kling-3, veo-3.1, veo-3.1-fast and veo-3.1-lite. The surcharge is $0.07 a second on kling-3, $0.25 on veo-3.1, and only $0.025 on both veo-3.1-fast and veo-3.1-lite. Per second of sound, the two cheaper Veo rows are ten times cheaper to add sound to than the standard Veo row.

All numbers are the per-video-second and per-video-second-audio rates in the Sume video catalog (read 2026-10-12). They can change, so check GET /v1/videos/models, the live catalog, before you commit a budget.

How much does sound add per second on each row?

The surcharge is the audio rate minus the silent rate. The multiple is audio divided by silent. A low surcharge in dollars is not always a low multiple, so both columns matter when you scale to hundreds of clips.

Arithmetic: kling-3 0.21 - 0.14 = 0.07 (0.21 / 0.14 = 1.5x). veo-3.1 0.50 - 0.25 = 0.25 (2.0x). veo-3.1-fast 0.125 - 0.10 = 0.025 (1.25x). veo-3.1-lite 0.0625 - 0.0375 = 0.025 (1.67x).

What does an 8-second clip with sound cost?

Veo rows run 4 to 8 seconds, so 8 seconds is their ceiling. Kling 3 runs 4 to 15 seconds, so it also appears at 15 seconds, a length the Veo rows cannot reach in one job.

Arithmetic for 8 seconds with sound: veo-3.1-lite 8 x 0.0625 = $0.50; veo-3.1-fast 8 x 0.125 = $1.00; kling-3 8 x 0.21 = $1.68; veo-3.1 8 x 0.50 = $4.00. For 15 seconds with sound, kling-3 is 15 x 0.21 = $3.15 versus $2.10 silent.

Which row should I pick when the clip needs sound?

Pick by what the sound must cost you, not by the surcharge alone. If sound is a nice extra on a draft, veo-3.1-lite at $0.50 for 8 seconds is the floor of this group. If you need 1080p or a clip longer than 8 seconds, only kling-3 qualifies here, because the Veo rows list 720p and 8 seconds as the maximum in the catalog. The standard veo-3.1 row is the only one where sound doubles the bill, so it is the row to run silent and add sound elsewhere if you must use it.

How do I request sound on one of these rows?

Send generate_audio in the POST /v1/videos body. The catalog default follows the model's audio capability, so set it explicitly when you are pricing a batch. Use an Idempotency-Key so a retry returns the original job instead of a second charge.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Idempotency-Key: veo-lite-sound-001" \
  -H "Content-Type: application/json" \
  -d '{"model":"veo-3.1-lite","prompt":"A barista slides a latte across the counter, cafe chatter","duration":8,"resolution":"720p","generate_audio":true}'

When is this the wrong comparison?

The surcharge only tells you what sound costs inside one row. It does not tell you which output you want. For that trade-off, read Kling 3 native audio surcharge vs adding narration separately.

• Rows with sound included in a single per-second rate, such as wan-3.0 at $0.0625/s at 480p, have no surcharge to compare, so rank them by total clip price instead.
• Some rows never go silent. Check the model entry for generate_audio before assuming you can switch it off.
• If you only need a narration over a silent clip, price a separate voice step against the sound surcharge.