OmniHuman2
OmniHuman 2.0 pricing starts at $0.07 per generated second for the film-grade tier, while BytePlus lists the lip-sync avatar tier at $0.12 per second. A 90-second video therefore works out to $6.30 or $10.80 at the posted rates, before any other costs. Budget for rejected takes too.
The two rates are for different jobs
The $0.07-per-second film-grade card emphasizes camera and body movement. It also mentions multi-person scenes and a 30-to-300-second output range. The $0.12-per-second avatar card emphasizes realistic talking avatars and lip sync from an image and audio file. These are separate offerings on the product page, not a base price plus an automatic lip-sync surcharge. Don’t add the rates together. Different tools, different jobs.
The distinction matters when you storyboard. If your shot calls for a person stepping through a room while the camera changes position, rather than a face held steady against a plain background, the film-grade description is the relevant starting point when you assign a budget to that scene. If the job is a direct-to-camera presenter reading a script with tight mouth movement, read the avatar description first. BytePlus doesn’t promise that every shot will fit either tier equally well; these are the uses its pricing cards actually describe.
The page also says “40% lower cost” for the film-grade option, but doesn’t identify the comparison baseline. I’d ignore that percentage in a purchasing spreadsheet. The dollar rate and the length of the finished shot are much more useful inputs.
What a finished video costs at posted rates
Multiply generated duration in seconds by the rate for the chosen tier. This is an estimate from BytePlus’s listed per-second prices, not a quote or an independently verified invoice. Here are useful planning sizes:
- 30-second film-grade shot: 30 s × $0.07 = $2.10. BytePlus describes 30 seconds as the low end of the film-grade output range.
- 60-second film-grade shot: 60 s × $0.07 = $4.20. Three attempts at that duration would be 3 × $4.20 = $12.60, assuming each attempt is billed at the posted rate.
- 90-second film-grade scene: 90 s × $0.07 = $6.30. Three complete attempts would be 3 × $6.30 = $18.90.
- 90-second avatar script: 90 s × $0.12 = $10.80 at the avatar rate. Three complete attempts would be 3 × $10.80 = $32.40.
- 300-second film-grade piece: 300 s × $0.07 = $21.00, using the top of BytePlus’s stated film-grade range. That assumes one successful generation. A multi-shot edit is different.
A 15-second social ad is a trap for quick mental math. The film-grade card specifies 30 to 300 seconds, so 15 s × $0.07 = $1.05 is not a supported film-grade job quote. You might trim a longer result in an editor, but you’d still need to check how BytePlus bills the generated duration. The product page does not give the avatar tier the same stated output range; don’t carry that minimum across by assumption.
Budget for takes, not just the approved cut
The difference between a cheap render and an affordable production is iteration. Say a 60-second film-grade scene needs one reference-image change and one revised audio take. If you generate the full minute each time, your planning figure is 3 × (60 s × $0.07) = $12.60 at the film-grade rate. One approved minute might cost more. This is a planning scenario. BytePlus does not specify a retry-billing policy on that page.
I would lock the spoken audio before generating a long presenter segment. Rewrite the script on paper, record a clean version, then check the reference photo before you start paying for full-length attempts. For image and audio preparation, the step-by-step usage guide covers the basic order of work; the reference-photo guide goes deeper on framing.
Keep a separate line in your budget for voice recording, editing, subtitles, revisions by a human, and any other service you use. The posted video-generation rate doesn’t price those jobs for you, even if the final video needs all of them before it can go in an ad, on a product page, or into a client review. Nor does it tell you whether a failed request, a canceled job, or a rerun is billable. Check the account terms first. The estimates are not ceilings.
What the public pricing page leaves open
The BytePlus product page advertises per-second prices but does not spell out a free trial, tax handling, volume discounts, minimum account spend, or the billing treatment of failed generations. It also does not publish the avatar tier’s output-duration bounds alongside its price. If a reseller gives you a different limit or bundled credit, that is the reseller’s offer, not a BytePlus platform rule.
For access details, BytePlus links to its OmniHuman 2.0 API document from the product page. Check the console’s current billing terms and ask BytePlus sales which tier applies and which generated seconds will appear on your invoice. Ask how failed requests appear too. If you need a quote for a campaign rather than a rough per-shot estimate, bring a shot list with durations to that conversation and ask whether rounding, account minimums, or retry billing changes the number you would actually pay.
For a quick account decision, first write down the runtime of each deliverable, then assign the tier based on the kind of motion you need. Add a realistic retake allowance. Compare that amount with your editing budget, not with the price of a single perfect render. The independent OmniHuman 2.0 overview explains the broader image-plus-audio workflow if you haven’t settled on the tool yet.
Before you approve the spend
Use a short shot list with one row per planned generation: source photo, audio filename, required motion, intended tier and duration. Add your budgeted attempt count. A row for a 90-second film-grade scene with two planned attempts reads 2 × (90 s × $0.07) = $12.60 at the posted film-grade rate. That’s more actionable than “video budget: $20.”
If the performance needs a camera move, make that explicit in the shot notes before you record audio. BytePlus says text prompts can control camera movement and actions, but it doesn’t guarantee your first prompt will produce the cut you want. Decide which motion is essential and which is a nice-to-have; revise the cheap part, the notes, before buying another long generation.
Keep the approved audio file next to the still image and write the runtime into the filename or project sheet. When the final voice edit grows by a few seconds, recalculate the line item before pressing generate. Reprice it. A few extra seconds still count in your plan.
