Seedance 2.0 Mini API and Pricing, Explained Through a Real Pipeline

Seedance 2.0 Mini is the smallest tier of ByteDance's Seedance 2.0 video model and the cheapest second in the family. It renders clips of 4 to 15 seconds with sound generated alongside the picture, from a text prompt, from an exact first frame, between a first and a last frame, or from reference images that carry a product or a character into a new scene. It reached OpenRouter in August 2026, and it joined Treza's video catalog on October 4.

This post covers how its video token price turns into a cost per second, how it compares with Seedance 2.0 and Seedance 2.0 Fast, and how to call Seedance 2.0 Mini as a published API endpoint. Every parameter and price below is read off Treza's live video model catalog as of October 2026, and every behavior described was checked with real renders.

What you get before you talk about price

Nothing in the input set is cut: prompt, keyframes at both ends, and reference images all work, and the clip arrives with its own sound. What Mini gives up is the top of the resolution range. The full Seedance 2.0 also renders 1080p and 4K; Mini stops at 720p, the same ceiling as Fast. Capability detail lives on the Seedance 2.0 Mini model page.

How Seedance 2.0 Mini bills

Seedance does not publish a flat rate per second. It bills video tokens, and the listed price for Seedance 2.0 Mini is $0.0000035 per token. The catalog lists the same price with audio on and with audio off, so switching the soundtrack off saves nothing.

ByteDance counts video tokens by frame area: width times height times 24 frames, divided by 1,024, for each second of output. On our launch week renders that formula landed within a fraction of a cent of the real charge every time, which makes the per-second cost easy to work out at a 16:9 frame:

So a 4 second draft at 480p is about 13 cents, a 10 second clip at 720p is about 76 cents, and the longest clip the model renders, 15 seconds at 720p, is about $1.13.

Because the meter is frame area, the shape of the frame moves the number a little. A 21:9 ultrawide frame carries more pixels than 16:9 at the same resolution label, so it costs more per second. Square and tall frames land close to the 16:9 figure.

Seedance 2.0 Mini vs Seedance 2.0 Fast and the full model

Three tiers of one model family, same inputs, three meters.

Per video token, Mini costs half of what the full Seedance 2.0 lists, and about a sixth less than Seedance 2.0 Fast. On a single clip the Mini-to-Fast gap is a few cents. It matters at volume: a hundred 4 second drafts at 480p run about $13 on Mini against about $16 on Fast, and a thousand of them is $134 against $161.

That arithmetic is what Mini is for. A shot that will be shown large belongs on Seedance 2.0 at 1080p or 4K. A finished vertical clip at 720p is Fast territory. Mini is the pick when the clip is one of many: hook variations, moving storyboards, the first pass on an idea before a final render on a larger model. When Mini and Fast look the same for your prompt, keep the cheaper one; when they do not, run it through both on two nodes and let the output decide. The Seedance 2.0 Fast pricing post covers the Fast side the same way.

For a clip longer than 15 seconds, step outside the 2.0 family. Seedance 2.5 renders up to 30 seconds in one clip at a higher rate.

Resolution and aspect: what came back

720p is both the ceiling and the default. Leave Resolution on Model default and a 9:16 clip comes back at 720 by 1280. Pick 480p and the same 9:16 clip comes back at 496 by 864, at 24 frames per second either way.

With a first frame attached and no aspect ratio set, the clip took the shape of the still: our square test image came back as a 640 by 640 clip at 480p. Set the aspect ratio explicitly whenever the still and the placement disagree.

In our launch week tests a 4 second clip took a little over a minute from submit to finished file at 480p from a prompt, and about a minute and a half with a first frame attached or at 720p. Mini is the cheapest tier, not a faster one.

First frames, last frames, and references

A first frame pins the opening image exactly, and the clip animates out of it. That is the technique the first frame post walks through, and the image to video generator covers it from the still's side. Add a last frame and the model fills in the motion between the two stills. On Mini that makes a storyboard cheap to animate: one still per beat, each the first frame of a 4 or 5 second clip.

Reference images do a different job. Instead of pinning a frame, they carry a subject into a scene it was never photographed in: the same backpack hanging in a mountain hut, the same character in a new location. Wire the stills into the Video Generation node's reference input and every one of them is sent to the model. Our reference probe is worth reporting honestly: its second test image pointed at a URL that did not exist, and the endpoint rejected the job at submit for exactly that image, before any rendering or any charge. A broken reference fails the run rather than being skipped, so check reference URLs before a batch.

ByteDance screens reference images and rejects recognizable real people before a job starts, so the run fails at submit rather than producing a clip. For photos of people, use Wan 2.7 or Grok Imagine instead.

About the audio

Sound is generated with the picture and the clip comes back with an AAC audio track, so a draft already plays the way the final will. Our cafe test prompt asked for ambient sound, and every clip came back with one.

For narration over the footage, add a Text to Speech node and mix it in with the Sequence node.

What you actually pay on Treza

Seedance 2.0 Mini through a Treza pipeline is metered from a prepaid credit balance. One credit equals $0.01, credit packs start at $5, credits never expire, and a failed generation charges nothing.

Before a run starts, the platform estimates the media cost of the graph and refuses to start one your balance cannot cover. For Seedance 2.0 Mini the estimator holds 8 cents of provider cost per second at 720p or with no resolution set, and 4 cents at 480p. Those are the per-second figures above rounded up to the cent, so the hold sits a little above what a 16:9 clip uses.

For a finished video costed end to end, narration and captions included, see the AI video cost breakdown.

Calling Seedance 2.0 Mini as an API without building the integration

Direct API access to a video model means credentials, a submit-and-poll job pattern, retries, moderation rejections, and somewhere to store the output.

On a Treza AI video pipeline, Seedance 2.0 Mini is a node on a canvas. Set the duration, aspect ratio, and resolution, wire a first frame, both frames, or reference images into it, place it next to a script, voiceover, or captioning step, and publish the whole chain as one versioned endpoint. Your application calls a typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route.

The swap is what pays off later. When the next model wins on price or quality, you change the dropdown and republish, and the endpoint your product calls does not change. Promoting a shot drafted on Mini to Fast or to the full Seedance 2.0 is the same one-field change. Our hosted MCP server exposes the same pipelines as tools, so an agent in Claude or ChatGPT can render a brief and return the link.

Where Seedance 2.0 Mini earns its place in a graph

Three jobs where the cheapest Seedance second does the work. Hook tests for ads, where a dozen 4 second openings at 480p cost under two dollars together and the winner gets re-rendered larger. Moving storyboards, where each beat of a script becomes a short clip from its own still. And short vertical cuts for a daily feed, drafted at 480p and rendered at 720p once the framing is right.

The fastest way to learn what it costs is to run your real prompt at your real duration and read the charge. Open the AI video generator, pick Seedance 2.0 Mini, and generate once.

Frequently Asked Questions

How much does Seedance 2.0 Mini cost per second?

The listed price is $0.0000035 per video token as of October 2026. At a 16:9 frame that works out to about 3.4 cents per second at 480p and about 7.6 cents per second at 720p. Audio on or off does not change the price. On Treza the charge comes from a prepaid credit balance, and a failed generation charges nothing.

How is Seedance 2.0 Mini different from Seedance 2.0?

It takes the same inputs, including native audio, first and last frames, and reference images, and bills half as much per video token. It renders at 480p or 720p, where the full Seedance 2.0 also offers 1080p and 4K.

What is the difference between Seedance 2.0 Mini and Seedance 2.0 Fast?

Both take the same inputs at 480p or 720p for 4 to 15 seconds, and both took about the same time to render in our tests. Mini lists $0.0000035 per video token against Fast's $0.0000042, so it is about a sixth cheaper per second: a few cents on one clip, real money on a batch of hundreds.

Does Seedance 2.0 Mini generate audio?

Yes. Sound is generated alongside the picture and comes back in the clip. For narration, add a Text to Speech node and mix it in with the Sequence node.

Can Seedance 2.0 Mini use reference images of real people?

No. ByteDance screens reference images and rejects recognizable real people before a job starts. For photos of people, try Wan 2.7 or Grok Imagine instead.

Can I call Seedance 2.0 Mini through an API without a ByteDance account?

Yes. On a Treza pipeline, Seedance 2.0 Mini is a node you configure once and publish as a versioned HTTP endpoint, with a typed /invoke route and an OpenAI-compatible route. Generations are metered from prepaid credits that start at $5 and never expire.