Seedance 2.0 Pricing: What It Actually Costs (API, Apps, and Credits)

The same 15 seconds, three very different bills

Seedance 2.0 pricing splits into two models, and knowing which one you are standing in decides your bill. Through the fal API you pay per second of generated video: $0.3034 at 720p with audio, $0.2419 on the fast endpoint, $0.682 at 1080p. Inside subscription apps you pay in credits: in Novoads, a 5-second clip is 3 credits and a full 15-second take is 7 credits, about $4.83 at the Pro plan rate. Run the same maximum-length clip through every door and it costs anywhere from $3.63 to $10.23. This guide maps every published price, works the 15-second example all the way through, and ends with the cheapest honest way to test the model. If what you are really asking is whether you can run it without paying at all, that is a different question with a different answer, covered in is Seedance 2.0 free.

Every way to pay for Seedance 2.0, in one table

Nobody sells Seedance 2.0 the same way twice. ByteDance built the model, API hosts meter it by the second, and ad platforms fold it into monthly credit plans. Before the deep dives, here is the whole market on one screen.

The one-table version

Route	Price shape	Best for
fal API, 720p with audio, standard tier	$0.3034 per second	Developers, standard quality
fal API, 720p fast tier	$0.2419 per second	High-volume production runs
fal API, 1080p with audio, standard tier	$0.682 per second	Hero cuts, big screens
ByteDance official (Dreamina app, cloud API)	No published Western price list	First-party access
InVideo subscription	Bundled plan credits	General video makers
Arcads subscription	Platform plans	Premium ad teams
Novoads credits	2.6 to 7 credits per clip	UGC-style ad testing

Every number in the fal rows is the published rate on the model's own fal page, re-checked on July 20, 2026. The app rows are deliberately not hard dollar figures: InVideo renders plan prices by region on its live page, and Arcads pitches the platform rather than a per-clip meter, so the honest cell is the pricing shape, not a number that may not match your screen. The ByteDance row is honest about a blank: both first-party doors are real, but neither publishes a simple Western rate card, which the ByteDance section below unpacks.

The two pricing models underneath

Everything above collapses into two philosophies:

• Metered (the API route): you pay for exactly what you generate. Right when volume is unpredictable and you have engineering time to spend.
• Subscription (the app route): a flat monthly amount buys a bundle of credits plus the workflow around the model. Right when the clip is a means to a finished ad.

Neither is universally cheaper. The rest of this post prices each one honestly, and the Seedance 2.5 explainer covers where the model family goes next.

The fal API: per-second rates for Seedance 2.0

The keyword most people search is "Seedance 2 API," and the API answer starts at fal, the host that serves the model to developers. The endpoint id is bytedance/seedance-2.0/text-to-video, with image-to-video and reference-to-video siblings, and its pricing page is refreshingly explicit.

The three published endpoints

fal lists three price points for Seedance 2.0, all per second of output:

• 720p with audio, standard tier: $0.3034 per second. The default door, maximum quality.
• 720p fast tier, with audio: $0.2419 per second. Lower latency and lower cost, positioned for production workloads, with no 1080p option.
• 1080p with audio, standard tier: $0.682 per second. The premium finish, at more than double the 720p rate.

There is also a Seedance 2.0 Mini on fal, but it sits in early access behind a request form ("This model is not yet generally available"), so it has no generally published rate yet. Watch that page if you are cost-sensitive: a mini tier is usually where per-second prices fall next.

What a clip costs at those rates

Per-second pricing turns every clip into simple multiplication. At the standard 720p rate:

• 4-second hook: about $1.21.
• 5-second test: about $1.52.
• 10-second demo: about $3.03.
• 15-second maximum take: about $4.55.

Switch to the fast tier and the 15-second ceiling drops to about $3.63. Switch to 1080p standard and the same 15 seconds jumps to about $10.23. That 2.25x gap between 720p and 1080p is the single biggest pricing decision on the API route, and it is not a positioning choice: the next section derives it from fal's own token formula. Most feed-first ad tests do not need the premium finish.

The fine print that changes your bill

Three spec details on the fal page quietly shape your spend:

• Duration: clips run 4 to 15 seconds, or "auto," where the model picks the optimal length for the content. An auto clip can bill more seconds than you pictured.
• Resolution: you choose the tier per request, so a 1080p habit compounds quietly across a whole test batch.
• Reference inputs: the reference-to-video endpoint accepts up to 9 images, 3 video clips, and 3 audio files in a single generation. Powerful for matching a brand look, and every second of the result meters at the same per-second rate.

None of this is hidden. It is just arithmetic that rewards people who read the schema before wiring money to it.

The per-second rate is a pixel meter in disguise

Below that price table, fal publishes something more useful than the rates themselves: the formula that produces them. It is worth a minute, because once you can reproduce a published rate with a calculator, the rest of the rate card stops looking arbitrary.

The formula sitting next to the price

fal charges $0.014 per 1,000 tokens for 480p, 720p and 1080p output, and it states what a token is: the output height multiplied by the width, multiplied by the duration, multiplied by 24, divided by 1,024. Run that on one second of 720p video at 16:9:

1. The frame is 1,280 by 720 pixels, which is 921,600 pixels.
2. Times 24 frames, divided by 1,024, gives 21,600 tokens for that second.
3. At $0.014 per 1,000 tokens, that second costs $0.3024.

That is the standard rate published on the image-to-video page, reproduced exactly. So the per-second price is not a second meter running alongside a token meter. It is the token meter, resolved at one fixed resolution and rounded into a unit a human can budget with.

It is also the cleanest answer to the sites promising unlimited generation at no charge, which the free Seedance question unpacks in full. The meter counts pixels, every additional pixel carries a marginal cost upstream, and there is no setting at which that cost reaches zero. Whoever is generating, somebody is paying per pixel.

Why 1080p bills 2.25 times more

Once the formula is visible, the biggest number in the table stops being a mystery. A 1080p frame is 1,920 by 1,080, which is 2.25 times the pixel count of a 720p frame. Divide the published 1080p standard rate of $0.682 per second by the published 720p standard rate of $0.3034 per second and you get 2.248.

The resolution premium is a pixel count, not a quality surcharge. That reorders where the savings actually are on the API route:

• Cutting a 10-second clip to 6 seconds saves 40% of the bill, every time, with no argument about quality.
• Dropping 1080p standard to 720p standard saves about 55%, and on a phone-sized placement most viewers could not tell you afterwards which one they watched.
• Reframing 16:9 to 9:16 saves nothing at all. A 720 by 1,280 vertical frame holds exactly the same pixels as a 1,280 by 720 horizontal one, so orientation is a placement decision with no money attached to it.

Quote the endpoint, not just the model

There is one more reason to name the tier out loud. The two Seedance 2.0 endpoint pages do not publish an identical standard rate. Text-to-video lists $0.3034 per second for 720p with audio. The image-to-video page's own pricing block lists $0.3024 per second for its standard endpoint, and works the example out to about $3.02 for 10 seconds.

A tenth of a cent per second reorganises nobody's budget. It matters for what it says about the shape of the thing: this is a per-endpoint rate card that can differ by fractions between surfaces, not one price attached to one model. Quote a Seedance rate with its tier and its endpoint attached, or do not quote a decimal at all.

What the per-second rate leaves out

The API rate is real, published, and fair. It is also not your cost per ad. Three gaps sit between $0.3034 per second and a creative you can actually run, and they are where API-first budgets quietly go wrong.

You pay for takes, not keepers

Generative video is iterative by nature. fal's own prompting guide for the model says it directly: "Iterate fast. Start with 5-second generations" to nail the style before committing to longer clips. That advice exists because first takes often miss: the camera move is wrong, the product warps, the pacing drags. Every miss bills at the same per-second rate as the keeper. If a scene takes three attempts before one is usable, your effective price per usable 15-second clip at 720p is not $4.55, it is closer to $13.65. Budget on attempts per keeper, not on seconds.

A clip is not an ad

Seedance 2.0 generates cinematic footage with native audio, meaning ambient sound and scene tone ship with the clip. What it does not ship is the rest of the ad: a scripted voiceover in your market's accent, burned-in captions, a hook structure, an actor presenting your product, aspect-ratio variants for each placement. On the API route each of those is another tool, another integration, and another line item. Our breakdown of video ad production costs walks the full stack; the short version is that the model is usually the cheapest layer in it.

The plumbing has a cost too

An API key is not a workflow. Someone has to write the integration, handle queue callbacks, store outputs, retry failures, and stitch results into an editable timeline. For a developer testing a product idea, that is a weekend. For a marketing team, it is the difference between shipping ads this week and owning an internal tool nobody asked for. This is a judgment call, not a published number, but it is the reason ad teams who can clearly do the math on per-second rates still end up inside apps.

The apps: Seedance 2.0 inside a subscription

The second way to buy Seedance 2.0 is to not buy Seedance 2.0 at all, but a subscription that includes it. Three doors matter here, and they price the same model three different ways.

ByteDance's own front door

ByteDance's Seed site is the model maker's official surface, and it references both Seedance generations while routing visitors to "Try Now" and "Get API" entry points rather than a consumer price list. There is no per-clip retail price to quote there: the page is a maker and developer surface, not a checkout. Priced access to Seedance 2.0 runs through hosted APIs like fal and the apps that license the model. Practical translation for a Western ad team: the maker's front door leads to developer access, and a per-clip price you can actually compare lives in the third-party products below.

InVideo: one model among 200+

InVideo folds Seedance 2.0 into a breadth play. Its paid plans include, in the pricing page's own words, "Access to 200+ image, video, audio, music models including Seedance 2.0, Veo 3.1, Kling 3.0," with exact plan prices rendered by region on the live page. The pitch is variety: one subscription, every frontier model, a general-purpose editor around them. The trade is that Seedance is one option in a very large menu, not the center of an ad-specific workflow.

Arcads: Seedance inside a premium ad platform

Arcads, an AI ad platform aimed at performance teams, currently banners "Seedance 2.0 is live" on its homepage and folds the model into its actor-led ad tooling. Read it as the premium end of the market: the homepage sells winning ads and a library of 1,000+ AI actors, not a per-clip meter, so you evaluate the platform first and the price follows. For a head-to-head on how this platform category compares, see the best AI video ad platforms roundup.

Seedance 2.0 in Novoads credits, exactly

Novoads is the third door, and since these are our own numbers, they can be exact. Seedance 2.0 is one of the video engines on the platform, priced per second in credits, and the schedule below is the live production number, not an estimate.

The per-second credit schedule

A Seedance 2.0 generation in Novoads runs 4 to 15 seconds and prices by length: 3 credits for 5 seconds, 5 credits for 10 seconds, 7 credits for the full 15-second take, with a 4-second minimum at 2.6 credits. The schedule is deliberately identical to Kling v3 Pro on the platform, so switching engines never changes your budget math; the Seedance vs Kling comparison covers when to switch anyway. Clips render at 480p or 720p, in six aspect ratios including 9:16 vertical, with native audio on by default.

What a credit costs on each plan

Credits translate to dollars through your plan, so the same 7-credit clip costs different money at different volumes:

• Inicial, $49 for 50 credits: $0.98 per credit, so 15 seconds costs about $6.86.
• Pro, $69 for 100 credits: $0.69 per credit, so 15 seconds costs about $4.83.
• Studio Light, $99 for 150 credits: $0.66 per credit, about $4.62 per 15-second clip.
• Studio Pro, $199 for 340 credits: about $0.59 per credit, about $4.10 per clip.
• Studio Max, $399 for 720 credits: about $0.55 per credit, about $3.88 per clip.

Notice what happens at the top of that list: at Studio Pro and above, the credit route drops below fal's standard 720p list rate for the same 15 seconds. Volume is the lever.

What the credits include

The credit price buys the generation inside a finished workflow:

• Write a script, or auto-generate one from your product and angle.
• Pick an AI actor who can hold and present your product on camera.
• Generate the Seedance clip, then add a voiceover in one of 30+ languages with real regional accents.
• Export vertical, square, or horizontal for each placement, from the same project.

That surrounding machinery is exactly the gap the API route leaves open, and it is why comparing raw per-second rates alone undersells what either side is for. For what the human alternative runs, UGC creator rates start around $50 per video and climb fast.

The 15-second worked example

Fifteen seconds is Seedance 2.0's maximum single take and the fairest benchmark, because it is long enough for a hook, a demo beat, and a payoff. Here is the same clip priced through every door.

The math, five ways

Route	Rate	15-second clip
fal, 720p fast tier	$0.2419 x 15	$3.63
fal, 720p with audio, standard	$0.3034 x 15	$4.55
Novoads, Studio Max rate	7 credits x $0.554	$3.88
Novoads, Pro rate	7 credits x $0.69	$4.83
fal, 1080p with audio, standard	$0.682 x 15	$10.23

Two routes bracket the market: the fast API endpoint is the floor at $3.63, and 1080p is the premium ceiling at nearly triple that.

The honest read

At 720p, the API and the credit plans land within about a dollar of each other, which means the decision is not really about the per-clip price. It is about what surrounds the clip. Choose by what you are actually buying: seconds of footage, or a finished ad with a script, a voice, and an actor attached. And remember the retry math from earlier applies everywhere: on any route, the team that needs two takes per keeper pays double the sticker price, which is why prompt quality is a cost lever. Our Seedance prompt guide for ads exists for exactly that reason.

The cheapest way to test Seedance 2.0

Testing is where most readers actually are: you do not need 500 clips, you need three good ones and a verdict. The cheapest path depends on which kind of builder you are.

If you write code first

Take fal's own advice and iterate at 5 seconds on the fast endpoint: about $1.21 per test clip. Ten test generations cost roughly $12, and you can graduate the survivors to the standard endpoint or 1080p for finals. This is the cheapest raw access to the model that exists at published rates today, and for a developer evaluating output quality it is the right door.

If you make ads first

If your endgame is a running ad rather than an API integration, test inside the workflow you would actually use. The Novoads trial is $1 for 3 days of access and grants roughly 10 credits: enough for three 5-second Seedance tests, or one full 15-second take plus a short retake, with the script, actor, and voiceover steps included. It is a paid trial, not a free plan, and it converts to $49/month on Inicial unless you cancel, which takes one click. Start for $1, make the clips, and judge the model on your own product instead of a showreel. If Veo is also on your shortlist, the Seedance vs Veo for ads breakdown prices that fork.

Use the API when, use an app when

• Use the fal API when: you have engineering time, unpredictable volume, or you are building Seedance into your own product and the clip itself is the deliverable.
• Use an app when: the deliverable is an ad, your volume fits a monthly plan, or the people running it are marketers who should be spending their week on angles and hooks, not queue callbacks.
• Use both when: you are big enough to prototype angles in an app and industrialize the winners through the API. The two doors are not rivals; they are stages.

Price the workflow, not the clip

Seedance 2.0 is cheap by any historical measure of video production: the most expensive route on this page delivers 15 seconds of finished, sound-on footage for about the price of lunch. That is exactly why the per-second rate is the least interesting number here. The clip was never the expensive part; the script, the voice, the actor, the retries, and the person assembling them were. So price the workflow, not the clip: pick the door that turns your next 10 attempts into a winning ad fastest, and let the model's few dollars disappear into the noise where they belong. If that door is an ad workflow, the $1 trial is the cheapest way through it.