Kling silent vs voiced: the credit math nobody publishes on our shelf

The silent-vs-voiced question hides inside every creator's budget and nobody publishes the answer in app credits. Kling's own pricing writes it in dollars (silent 720p $0.084/s, voiced $0.112/s as of 2026-10-20) and leaves the user to work out whether a silent render with a voiceover added in post is cheaper than a voiced render in-pass. On our shelf the two numbers sit next to each other in lib/models.ts and the arithmetic is three lines.

The rate card

Kling O3 Pro on our shelf, 720p (the one resolution Kling O3 Pro serves):

audio	credits per second	¢ on Lite	¢ on Creator	¢ on Studio annual
voiced	30	30.0 ¢	22.5 ¢	16.5 ¢
silent	24	24.0 ¢	18.0 ¢	13.2 ¢

The delta is 6 credits a second, or 20% off the voiced rate. In dollars that is 6¢/s on Lite, 4.5¢/s on Creator, 3.3¢/s on Studio annual. Multiply by a scene length and you get a number worth knowing before you flip the audio toggle.

Compare against the two other voiced-capable engines on our shelf:

engine	720p voiced	720p silent	delta
Kling O3 Pro	30 cr/s	24 cr/s	6 cr/s (20%)
Veo 3.1 Fast	30 cr/s	20 cr/s	10 cr/s (33%)
Seedance 2.5	70 cr/s	70 cr/s	0 (ByteDance bills by resolution, not by audio toggle)
Seedance 2.0	45 cr/s	45 cr/s	0
Seedance 2.0 Fast	50 cr/s	50 cr/s	0

Two important facts the SERP cost guides keep burying. First, Seedance never discounts for silent — the engine prices by resolution, so turning off the audio toggle on a Seedance render does not change the credit meter. Second, Veo's silent discount is bigger than Kling's: 33% vs 20%. If the shot is 8 seconds or under and does not need native audio, Veo silent 720p at 20 cr/s is the cheapest voiced-engine second on the shelf, below even Kling silent at 24.

The 30-second b-roll build

Here is the brief: 30 seconds of b-roll under a voiceover. The voiceover rides on top in the editor; the renders themselves do not need to generate audio in-pass. Six 5-second shots, cut on the beat.

Build A — Kling silent + VO: 6 × 5s at 24 cr/s silent 720p = 720 credits. The character carries across all six shots because @Element1 is the same reference set each time.

Build B — Kling voiced in-pass (no VO on top): 6 × 5s at 30 cr/s voiced 720p = 900 credits. The engine generates audio for every second. If the brief does not use that audio (you muted it and dropped a voiceover anyway), you paid 180 credits for a track that goes in the trash.

Build C — Veo silent 720p + VO: 6 × 5s at 20 cr/s silent = 600 credits. The 8-second Veo cap is not hit because every shot is 5s. The cheapest same-length build.

Build D — Seedance 2.5 silent 720p + VO: 6 × 5s at 70 cr/s = 2,100 credits. Seedance does not refund for silent, so the shelf's priciest-per-second engine stays priciest. You pick Seedance here only if the shot needs its 15–30s one-take range or 30 references.

build	credits	$ on Lite	$ on Creator	$ on Studio annual	notes
A — Kling silent + VO	720	$7.20	$5.40	$3.96	the character-first b-roll
B — Kling voiced in-pass	900	$9.00	$6.75	$4.95	the in-pass audio you may not use
C — Veo silent + VO	600	$6.00	$4.50	$3.30	the cheapest same-length build
D — Seedance 2.5 silent + VO	2,100	$21.00	$15.75	$11.55	the one-take or 6+ character row

Build A costs 180 credits less than Build B — that is the exact silent discount on Kling, applied across the scene. On a daily cadence of five 30-second b-roll pieces per week over four weeks, that is 180 × 20 = 3,600 credits saved per month, which is about $27 on Lite ($10/1,000) or $20.25 on Creator ($45/6,000) — enough to render a sixth piece every week.

When silent-first is the wrong answer

Three cases where you flip the toggle back:

Lip sync. The character is speaking a line, the mouth has to match. Kling's voiced path renders the lip motion against the audio in-pass; a silent render with a voiceover added in post needs a separate lip-sync pass, which eats time and often another engine's credits. The 20% silent discount is less than the cost of a bad sync. Flip to voiced.

On-screen diegetic audio. The character says one line that is diegetic to the scene — a door slam, a footstep, a laugh that has to come from the body. Veo 3.1 Fast is the row here, native audio is tight (public writing on Veo 3.1 as of 2026-10-20 puts dialogue sync at around 120ms), and the 10 cr/s delta is paid for the in-pass audio that would be hard to fake.

The reference-mode length cap on Veo. A reference render on Veo is always 8 seconds. If your brief needed 6s and you tried to save credits by going silent at 6s, Veo will still render 8 and bill 8 (160 cr voiced, 160 silent at 1080p — same price again because Fast tier holds price across resolution). The silent discount on Veo 720p at 8s is 10 cr/s × 8s = 80 credits — still worth it if you can use the 8s, worthless if your edit calls for 6s.

The silent-first b-roll cadence we run

Our own 9.7 million views in 30 days on @cevrico came from a format that happens to be silent-first b-roll under a voiceover. The template, with the credit math:

step	engine	per piece	per month (20 pieces)	$ on Creator
cold open 4s	Kling O3 Pro silent 720p	96 cr	1,920 cr	$14.40
b-roll 20s (4 × 5s)	Kling O3 Pro silent 720p	480 cr	9,600 cr	$72.00
close-out 4s	Kling O3 Pro silent 720p	96 cr	1,920 cr	$14.40
voiceover added in editor	—	0 cr	0 cr	—
finished 28s piece	—	672 cr	13,440 cr	$100.80

Creator at $45 covers 6,000 credits — one week of this cadence at a 20-piece/mo rate uses about 3,360 credits, so the plan sustains a monthly cadence without a top-up. If the same 20 pieces were rendered voiced in-pass on Kling, the credit meter rises to 16,800 cr/month ($126 on Creator) and the voiceover you added anyway sits on top of an unused native audio track.

Own results — your numbers depend on what you render, how long the pieces are, and whether the b-roll format fits the brand. Earnings projections are not a thing we publish; what we publish is the credit meter, which is the thing a plan has to cover.

The decision rule

Shot has dialogue the mouth has to match: voiced, Kling O3 Pro.

Shot is 8 seconds or under and needs native audio (a line, a door slam, diegetic sound): voiced, Veo 3.1 Fast — the 33% silent discount is bigger here but the engine's native audio is the reason to pick it.

Shot is b-roll under a voiceover (no diegetic audio needed): silent, Kling O3 Pro at 24 cr/s or Veo at 20 cr/s silent.

Shot is a Seedance render: audio toggle does nothing to the credit meter. Pick for shot length or references, not for the audio toggle.

Lite ($10 / 1,000 credits) runs about 42 silent Kling seconds a month at this cadence, Creator runs 250, Studio 625. A free account's 300 credits is enough for a 12-second silent Kling test, which is three cuts of your own b-roll under a scratch voiceover. Flip the audio toggle on your own scene and see what the delta looks like against your brief.

The 20% delta is a lever creators under-use. Nobody publishes it because nobody carries both engines on one credit meter — the silent-vs-voiced number is a shelf fact, not a provider fact.