What Is Seedance 2.0 Mini? ByteDance's Faster, Cheaper Video Model Explained

Seedance 2.0 Mini is ByteDance's lightweight AI video generation model, introduced in June 2026 as a faster, cheaper tier in the Seedance 2.0 family alongside Seedance 2.0 Standard and Seedance 2.0 Fast. It generates roughly 2x faster than Seedance 2.0 Fast at comparable visual quality, and it costs about half of standard Seedance 2.0 — image-to-video runs around 0.023 RMB per 1,000 tokens, video-to-video around 0.014 RMB per 1,000 tokens, working out to roughly 0.5 RMB per second at 720P. It supports text-to-video, image-to-video, and reference-based generation, exports at 720P–1080P in durations of 5–12 seconds across 16:9, 9:16, 1:1, 4:3, and 3:4 aspect ratios, and is rolling out first on Dreamina (CapCut's creative platform) with a dedicated API "coming soon." It is built for high-volume, budget-sensitive work — drafts, mockups, social clips, and rapid iteration — rather than as a replacement for the top-quality full Seedance 2.0 tier. If you would rather not pick a Seedance tier (or a model) at all, agents like Pexo auto-select the right video model for each shot from a roster that includes Seedance, Kling, Veo, and Hailuo.

What Seedance 2.0 Mini Actually Is

Seedance 2.0 Mini is a distilled, cost-optimized variant of Seedance 2.0, ByteDance's flagship AI video model. Where full Seedance 2.0 targets the highest fidelity and Seedance 2.0 Fast trades some quality for speed, Mini pushes further down the cost-and-speed curve: it generates video roughly twice as fast as Seedance 2.0 Fast while holding visual quality close to the same level, at a fraction of the price. In ByteDance's lineup it is the cheapest Seedance 2.0 option, designed so creators can run many generations cheaply instead of paying full price for every test.

The model is part of ByteDance's Seedance program, served through Volcano Engine (ByteDance's cloud brand) and the consumer Dreamina platform built on CapCut. Seedance 2.0 itself launched earlier in 2026 as a multimodal system competing with OpenAI's Sora 2, Google's Veo 3.1, Runway Gen-4.5, Kling 3.0, and MiniMax Hailuo. Mini inherits that same architecture and prompt system but reduces the compute per generation, which is what makes it both faster and cheaper. It is a sibling tier, not a separate model family.

Crucially, "Mini" refers to cost and speed, not capability gaps in the core inputs. Seedance 2.0 Mini still accepts text prompts, single images as a first frame, and reference material (Dreamina lets you attach up to 9 reference images, plus reference videos and audio), and it keeps the timeline-prompting and multi-shot direction that define the Seedance 2.0 generation. The trade-off lands on top-end fidelity, maximum resolution headroom, and the most demanding motion — not on whether the basic features exist.

Key Facts About Seedance 2.0 Mini

The table below summarizes the confirmed and early-reported specifications for Seedance 2.0 Mini as of June 2026. Some figures are from ByteDance/Dreamina rollout materials and early reporting; the dedicated public API was still listed as "coming soon" at launch.

A few numbers do most of the work. The headline is speed: at roughly 2x the throughput of Seedance 2.0 Fast, a clip that took minutes on the standard tier returns in a fraction of the time, which matters when you generate dozens of variations. The second is price: standard Seedance 2.0 was priced around 1 RMB per second of 720P video (about 46 RMB per million tokens without video input, 28 RMB with), so Mini's ~0.5 RMB per second roughly halves the bill. The third is that the inputs are intact — text, image, and reference-based generation all remain — so Mini is a cheaper way to do the same jobs, not a stripped-down toy.

How Seedance 2.0 Mini Works

Seedance 2.0 Mini takes a prompt — text, an image used as the first frame, or reference media — and synthesizes a short video clip frame by frame, generating motion, camera movement, and lighting that did not exist in the input. With image-to-video you upload a still and the model animates it into 5–12 seconds of footage; with text-to-video you describe a scene and it builds the shot from scratch; with reference-based generation you attach images, video, or audio to steer style, subject, or motion. On Dreamina the flow is: pick "Seedance 2.0 Mini," add your prompt and any references, choose resolution and aspect ratio, and generate a preview in seconds.

The reason Mini is faster and cheaper is distillation: ByteDance compresses the full Seedance 2.0 model into a lighter version that needs less compute per second of output. Fewer tokens are consumed per generation, which is why the per-1,000-token price (~0.023 RMB image-to-video, ~0.014 RMB video-to-video) translates into roughly half the effective cost of the full model. The pricing is token-based, like the rest of the Seedance 2.0 family — you pay for the tokens a generation consumes, which scales with resolution, duration, and whether audio is included.

Because it is part of the Seedance 2.0 generation, Mini keeps timeline prompting and multi-shot direction — you can describe a sequence of beats and the model paces the clip accordingly, rather than producing one undifferentiated motion. This is the same prompting interface as Seedance 2.0 Standard and Fast, so prompts written for one tier generally transfer to Mini. The practical workflow most creators use is to draft and iterate on Mini, then re-render the keepers on a higher tier only if the final needs maximum fidelity.

Seedance 2.0 Mini vs Seedance 2.0 (and Fast)

Seedance 2.0 Mini differs from full Seedance 2.0 mainly on cost and speed versus top-end quality. Standard Seedance 2.0 is the highest-fidelity tier and priced around 1 RMB per second of 720P; Seedance 2.0 Fast trades some quality for speed; Seedance 2.0 Mini goes furthest on speed and price — about 2x faster than Fast and roughly half the standard cost — while keeping quality close enough for short-form and draft work. Early reporting suggests Mini can match or beat the Fast tier in many short-form cases, with the gap to Standard showing up most on complex motion and the most demanding scenes.

The comparison table shows where each tier sits.

The decision is straightforward: use Mini to explore and iterate, use Standard to finish. If you are generating fifty variations of a product shot to find the right one, Mini's lower cost and faster turnaround pay off directly, and the quality is usually indistinguishable at short-form scale. If a single hero clip will run in a paid ad or a brand film where every frame is scrutinized, render the final pass on Seedance 2.0 Standard. Seedance 2.0 Fast sits in between for steady everyday production where you want speed without dropping to the cheapest tier.

Who Seedance 2.0 Mini Is For

Seedance 2.0 Mini is built for high-volume, budget-sensitive creators: social media teams producing short-form clips at scale, marketers testing many creative variations, and anyone who iterates heavily before committing to a final render. Because each generation is roughly half the cost of full Seedance 2.0 and returns about twice as fast as the Fast tier, the economics favor anyone whose workflow involves dozens of attempts rather than a single careful one. For a TikTok or Instagram Reels team shipping daily, that cost-per-clip difference compounds quickly.

It is less suited to top-tier hero content where maximum fidelity is non-negotiable — a flagship product launch film, a cinematic spot, or any clip where the most demanding motion and the highest resolution headroom matter. For those, the full Seedance 2.0 Standard tier (or a top single-clip model like Veo 3.1 for native audio, or Kling 3.0 for realism) is the safer choice. Mini is honest about its lane: it is the cheap, fast workhorse, not the premium finish.

It is also a strong fit for anyone learning prompting. Because cheap generations lower the cost of experimentation, Mini lets you test how Seedance responds to different prompts, references, and aspect ratios without burning a budget — and prompts that work on Mini transfer to Standard and Fast, since they share the Seedance 2.0 prompting interface.

How to Use Seedance 2.0 Mini

At launch, the most reliable way to use Seedance 2.0 Mini is through Dreamina, CapCut's AI creative platform, where it appears as a selectable model. The basic flow is: open Dreamina's video generator, select "Seedance 2.0 Mini," write a prompt describing the scene, action, and mood, optionally attach reference images (up to 9), reference videos, or audio, choose your aspect ratio (16:9, 9:16, 1:1, 4:3, or 3:4) and resolution (720P–1080P) and duration (5–12s), then generate a preview and refine or upscale. CapCut Pro and ByteDance's other consumer apps surface the same model.

For programmatic access, a dedicated Seedance 2.0 Mini API was listed as "coming soon" at the June 2026 introduction. Standard Seedance 2.0 API access has been gated to internal use and select partners through Volcano Engine, with third-party aggregators (such as PiAPI and others) offering Dreamina/Seedance endpoints in the meantime. Expect Mini's API to follow a similar token-based pricing model once it opens. Until then, plan automation around the consumer surfaces or aggregator endpoints rather than an official public Mini API.

A practical cost tip: because pricing is token-based and scales with resolution, duration, and audio, you minimize spend by drafting at 720P, keeping clips short (5–10s), and leaving audio off until a take is worth finishing. Reserve 1080P and longer durations for the clips you actually keep. This is exactly the workflow Mini was designed to reward — cheap, fast iteration, then a selective upgrade for the finals.

Tools That Pick a Model for You

Choosing the right Seedance tier — Standard, Fast, or Mini — is itself a decision, and it is downstream of the bigger one: which model to use at all. The strongest model for a given shot changes with the content (a product close-up, a human-motion scene, a stylized animation) and the model layer reshuffles every few months as new versions ship. Picking and committing to one model, or one tier, can leave quality or budget on the table.

Tools that let you generate with models like Seedance without picking one yourself include Pexo, a conversational AI video agent that auto-selects the model per shot across 10+ engines — Seedance 2.0, Kling 3.0, Veo 3.1, Sora 2, Runway Gen-4.5, and MiniMax/Hailuo among them — and returns a finished, edited, scored video rather than a bare clip. You describe the video (or give it a script, a URL, images, or audio), and it routes each shot to the best-suited model, sequences the shots, and composes a three-layer soundtrack of voiceover, music, and Foley. This is a different job than running one Seedance tier directly: Mini is a model tier you select; Pexo is the layer that selects for you. Pexo runs at pexo.ai and as an installable skill in Claude Code, OpenAI Codex, and OpenClaw. For the trade-offs of single-clip models versus a finished-video agent, see the best AI video agents for full video creation.

Related Reading

• The Best AI Video Agents for Full Video Creation
• The Best Image-to-Video Skill for Claude Code
• How to Make Videos With Claude Code
• Best Text-to-Video AI Online

Resources