Does Wan 3.0 cost less with the audio off? Same price on Sume

No. Wan 3.0 on Sume is priced per output second and per resolution, and the generate_audio flag is not part of the price: an 8-second 720p clip is $1.00 with audio on and $1.00 with audio off. If you want a cheaper clip, lower the resolution or the duration, not the sound switch.

Sume lists Wan 3.0 as wan-3.0 with a 2 to 30 second range (Video generation). This page shows what the audio field does and does not do to your bill.

The price formula has no audio term

Sume bills the provider list rate multiplied by 1.25 on every video model (Video Router). The figures here follow Sume's price estimator for wan-3.0. The live numbers for your workspace are in the catalog response and on the API pricing page.

For Wan 3.0 the formula is seconds multiplied by one rate per resolution: $0.0625 a second at 480p, $0.125 at 720p and $0.25 at 1080p. The estimator accepts generate_audio as an input, records it on the quote, and leaves the amount unchanged. The same 8-second 720p request with either value gives 1,000,000 USD micros, which is $1.00.

What actually moves the price

Two inputs change the amount for Wan 3.0: the duration and the resolution. Each extra second adds one second of the chosen rate. A 30-second 1080p clip is $7.50, and a 30-second 480p clip is $1.88 (the exact figure is $1.875).

• Resolution: 1080p costs four times 480p per second.
• Duration: linear from the 2-second minimum.
• Audio: no effect on the amount, so leave it at the model default unless you need silence for a mix.
• Retakes: a new take is a new job at the full price, so a prompt that needs three tries costs three times.

Reserve and capture still apply

Sume reserves the estimated amount when it accepts a job, captures it on success, and refunds the reservation if the job fails or is cancelled before capture (Usage, Generation admission). So the balance you need on hand at submit time is the price of everything in flight, not only the final total.

Because audio does not change the quote, the reservation for an 8-second 720p Wan clip is $1.00 whether you ask for sound or not. Read usage.cost on the completed job to confirm what was captured; the cost field is the Sume billable amount.

Check it yourself before a batch

Submit two single jobs with different generate_audio values and compare usage.cost in the poll response, or read GET /v1/usage?job_id= for each. If the two match, you can drop the flag from your cost model. Models differ on this point, so do not copy the result to another model without testing; each catalog row lists its own capabilities and prices through GET /v1/videos/models.

The practical rule: budget Wan 3.0 as seconds times a resolution rate, and spend your savings decisions on resolution and length.

GET /v1/balance returns the USD balance if you want to see the effect of a first job.

curl https://api.sume.com/v1/balance \
  -H "Authorization: Bearer $SUME_API_KEY"