Kling 3 sound on or off: $0.07 a second extra, when it pays
On Sume, Kling 3 costs $0.14 a second silent and $0.21 with sound. Compare the 5 to 15 second totals with scoring the clip yourself.

Turning sound on for kling-3 adds $0.07 per second on Sume: $0.14 silent, $0.21 with audio. A 10-second clip is $1.40 or $2.10. The flag is generate_audio, and the catalog default is on, so you only save money by sending false.
The difference per clip
The extra is linear in length, so the premium for sound on a 15-second clip is three times the premium on a 5-second one.
| Length | Silent | With sound | Sound premium |
|---|---|---|---|
| 5 s | $0.70 | $1.05 | $0.35 |
| 10 s | $1.40 | $2.10 | $0.70 |
| 15 s | $2.10 | $3.15 | $1.05 |
Silent plus your own bed
If you were going to replace the sound anyway, compare two routes. A 10-second silent clip is $1.40. A music generation is $0.125 per audio and a Timeline render is $0.10 per output minute (rounded up), so a silent clip with a music bed and a Timeline join lands near $1.63, against $2.10 for the clip with built-in sound.
The silent route is about $0.47 cheaper in that case, at the price of two more jobs to track. It also wins when you want to control the track: your own music, your own voice, or a licensed bed. If the clip would go out with its built-in sound as it is, the single job at $2.10 is simpler.
Set it explicitly
Send the flag every time so a default never changes your bill.
{
"model": "kling-3",
"prompt": "Hands wrap a gift, close-up, warm lamp light",
"duration": 10,
"generate_audio": false
}A rule of thumb
Leave sound on when the clip is going out as it is: a social post, a product loop with ambient noise, a draft you want to judge as a viewer would. Turn it off when a voiceover, a licensed track or a platform sound will replace it, because you would pay for audio you discard.
Sume reserves the full price at submit, so the clip price must be free in the balance when you send the request. If it is not, the call fails with 402 insufficient_credits before any provider work starts.
- Testing prompts: silent is enough, and 33% cheaper per clip.
- Final render with ambience: sound on.
- Final render under a voiceover: silent.
Sources
Related posts
More in Pricing
- MAI-Transcribe-2 prices end with the year: budget 500 hours a month
Both MAI-Transcribe-2 prices are limited-time through year end. 500 hours a month is $50 batch, $270 streaming, $300 on Sume. Microsoft states no 2027 price.
- MiniMax H3 Max 1080p costs twice 768p: the three tiers priced
H3 Max (minimax-h3-max) on Sume: 480p $0.0625, 768p $0.10, 1080p $0.20 per second, 5 to 15 seconds. 1080p is refined from native 768p.
- MiniMax H3 price table: 480p, 768p, 2K, 4K for 5, 10, 15 s
What MiniMax H3 (minimax-h3) costs on Sume per clip: 5, 10 and 15 seconds at native 480p and 768p, plus the 2K and 4K upscales. No 720p.
- Nano Banana 2.1 on fal bills tokens: $35.294 per million
fal bills Nano Banana 2.1 image output at $35.294 per million tokens, about $0.040 at 1K. Sume uses a flat per-tier card. Where to read your price.
Written by Sume