Credits per second, not USD per second
Video on Versely meters in credits per second. Native audio is a rate toggle. A host's dollars per second are not this catalog's price.
Video on Versely meters in credits per second. Native audio is a rate toggle on the rows that price it. A host's dollars per second are that host's price, not this catalog's.
The industry habit is to compare models in dollars per second, because that is how API hosts publish. A studio that meters many vendors through one balance cannot repeat those figures: they go stale the next time a host changes a cell, and they would not match the invoice. The honest unit, on this catalog, is credits per second of a named row, at the options you actually selected. The cost hub is the worked arithmetic. Billing type is the name of the meter. This page is the trend: think in that unit, or you will budget next week in a currency the invoice will not use.
This is not a 4K surcharge explainer. Resolution-banded rows are a lookup of their own. This is not "sound doubles Veo." That toggle has its own page. This is the unit of thought for a week of video.
The unit is a second on a named row
Billing type is which measurable thing a model's price is keyed to. Per-second billing charges more the longer the clip runs. Flat billing charges the same credits for the job regardless of length. Other rows key off resolution, megapixels, or characters. None of that is a setting you choose on a given model. It is how that row is published.
For video, the dominant meter on this catalog is still time. credits = rate_per_second(options) × seconds. The rate is not "video." The rate is this slug, at this resolution band or this audio setting, for this duration ladder. Seedance 2.0 is resolution_based_per_second: SD, HD, and 4K are different credits-per-second cells (11 / 25 / 219 on the current matrix), times a duration from 4 through 15 seconds. Veo 3.1 is an audio-dimension meter: 16 credits a second silent, 32 with native audio, times 4 / 6 / 8 seconds. Gemini Omni Flash Edit is 10 credits per second of the source clip, every edit turn.
Those numbers are catalog facts. They will move when the catalog moves. Read the card before you confirm. The app quotes the whole-credit total. A spreadsheet that hard-codes last month's rate is how a "cheap" week becomes a surprise.
Named row means the slug, not the press name. "Kling 3.0" is a market sentence. Several Kling rows exist, with different meters. "Grok Imagine" is stills, video, and extend. Budgeting "ten seconds of Kling" is not a number. Budgeting ten seconds on a specific slug, at a specific band, is.
Duration ladders are part of the unit. You cannot buy 7 seconds on a row that only sells 4 / 6 / 8. You buy 8 and trim, or you pick a row that exposes the seventh second. The unused second is still billed if the meter is per-second and the ladder forced you up. That is not a surcharge. It is the ladder.
Retries are part of the week. A per-second row does not give you a free miss. Five eight-second tries are forty seconds. Silent drafts, where the row allows a mute rate, are how you stop paying the sounded rate on every miss. If the test is the sound, mute drafts teach you nothing.
Images are usually not this meter. Nano Banana 2 Lite is a flat per-call still in the catalog data. Do not divide a still's credits by a fictional duration to make a "per-second" comparison with video. Different billing types are not one number.
Audio-native is a rate toggle
Native audio means the video model generates soundtrack — dialogue, effects, ambience — in the same pass as the picture. On some rows that capability is also a priced switch: the per-second rate is replaced when audio is on.
On Veo 3.1 the silent rate is 16 credits a second and the sounded rate is 32. The formula on AI video with sound is credits = rate_per_second(audio on | audio off) × seconds. Flip the toggle and every second of the clip moves to the other rate. There is no line item for "the talking part." A clip that is silent for five seconds and voiced for three still bills eight seconds at the audio-on rate if the toggle is on.
That is the trend that matters for budgeting, not the creative argument. Native audio is winning as a default on the models that matter this quarter — Seedance 2.0, Veo 3.1, Kling 3.0 all produce synced audio in one pass (Magic Hour; wavespeed). The bill only follows that default on rows that expose audio as a dimension. Other rows bake sound in, or have no sound, or want a separate TTS attach.
Additive voiceover is a different meter: mute (or silent) video + speech billed per 1,000 characters + a flat attach. It does not replace the per-second video rate. It sits next to it. Use it when the picture should be mute and the read should be reusable. Use the toggle when the mouth and the room have to be born with the frame.
Do not turn native audio on "just for the line" and expect mute pricing on the pads. Do not add native audio and a full voiceover on the same take unless you meant to pay twice for sound.
Seedance's matrix on this catalog is resolution-banded, not an audio pair. Do not assume Veo's 2× toggle is "how native audio costs." Assume nothing. Read the row.
USD-per-second is the wrong quote
Hosts publish dollars per second because they sell one model, or one family, as an API. Those figures are real for people who pay that host. They are the wrong quote inside a studio that resells many families through credits.
Do not:
- Take a third-party Seedance or Omni Flash dollar rate and divide it into Versely credits to "check" the margin. That is not a public number, and it will not match the invoice.
- Compare two Versely rows by converting each to dollars at a plan's credit bundle and calling that "USD per second." Plans change. Bundles are not a unit cost of a second.
- Paste a host blog's $0.067/s or $0.10/s into a client deck as "what this will cost in Versely." Those are someone else's SKU.
Do:
- Quote credits for the job: this slug, this duration, this toggle, this band, times the number of takes you actually expect.
- Quote a week as a sum of those jobs, plus captions, plus stills that are not per-second.
- When a client insists on a currency, quote the credit purchase, not a fabricated per-second dollar.
Third-party USD belongs in an explainer only with the vendor's name on it, and never as a Versely unit cost. Google's Omni Flash $0.10/s at 720p, Google's Lite $0.000034/image — those sentences are Google's. They do not appear on this page as a studio price, and they should not appear in your rate card either.
The trend underneath the pedantry is simple. Compute for video is still expensive, and it is still sold by the second. A studio that wants one bill has to pick a unit that is not each vendor's blog. Credits per second is that unit. It is worse for headlines. It is better for weeks.
Budget the week in credits
A useful week is a list of jobs, not a vibe about "AI video being cheaper now."
Write the clips. For each one: slug, seconds, audio on or off if the row toggles, resolution band if the row bands, expected retries. Multiply. Add stills as calls. Add captions as their own meter (captions on a month of clips is the worked version of that line). The sum is the week. If it does not fit the balance, you do not have a cheaper model problem. You have a too-many-sounded-seconds problem, or a too-many-retries problem, or a 4K-on-a-steep-band problem.
Draft mute when the row allows it and the test is the picture. Pay the sounded rate when the test is the line. Trim before Flash Edit, because Flash bills the source length every turn. Do not generate 15 seconds to use 6 on a per-second row.
Failover does not change the unit. If Sora's API is gone on 24 September, the next row still bills credits per second (or per call, or per character). The map of jobs is how you know which rate you just inherited. The unit stays credits.
When the next estimate should be a credit total on a named row rather than a dollar-per-second guess, sign up and read the quote the app shows before you confirm.
FAQ
Why won't Versely publish a USD-per-second rate next to each model?
Because the catalog is credits, and a dollar figure next to a row would be a unit cost this site is not in the business of stating. Hosts can publish their own dollars. The cost hub publishes credit formulas.
Is native audio always twice as many credits?
No. On Veo 3.1 the sounded rate is 32 credits a second and the silent rate is 16, which happens to be 2×. Other audio-dimension rows use different pairs. Some native-audio models do not expose a mute toggle at all. Read the matrix on the slug you are running.
Can I compare Seedance and Veo by converting both to dollars?
Not usefully, and not as a Versely price. Compare credits for the job you will actually run: duration, band, audio setting, retries. A 15-second Seedance HD pass and an 8-second Veo sounded pass are different objects. Making them one dollar number hides that.