Built by ByteDance · Accessible internationally via official API

Seedance 2.0 Official Website

If you searched for the Seedance 2.0 official website, you probably want to find the real source, log in safely, or start creating. Below: every official ByteDance channel, the full model card, and how to access Seedance 2.0 from anywhere in the world.

By Jay Yang·AI Video Technology·8 min read·

Key facts at a glance

Vendor

ByteDance

Brand family

Doubao (豆包)

Model line

Seedance

Version

2.0

Released

Feb 12, 2026

API public beta

Apr 2, 2026

Predecessor

Seedance 1.5 Pro (late 2025)

Modality

Text-to-video, Image-to-video

Max duration

12 seconds (consumer)

Max resolution

1080p

Aspect ratios

16:9 · 9:16 · 1:1

Native audio

Yes (visual + sync audio)

Sources: ByteDance Seed launch post (Feb 12, 2026), Volcengine Ark documentation, Wikipedia. Verified May 2026.

TL;DR

  • Seedance 2.0 is the current main-line video model from ByteDance, released February 12, 2026; the public-beta API on Volcengine Ark opened April 2, 2026.
  • There is no standalone English consumer site for Seedance 2.0 — distribution runs through three official channels: Volcengine Ark (developers), Doubao (Chinese consumers), and Jimeng (Chinese creators).
  • Output spec on consumer tier: 4-15 second clips up to 2K (native 480p/720p), with native synchronized stereo audio (BGM, SFX, voiceover, lip-sync) and multi-reference image input.
  • For global creators without Volcengine credentials or a Chinese phone number, third-party hosted interfaces wrap the official Seedance 2.0 API in an English UI with international payments.

(as of May 2026)

Plain language

When people search for the "Seedance 2.0 official website" they usually mean one of three things: where ByteDance publishes it, how to access the model, or how to log in. The honest answer is that ByteDance does not operate a standalone English consumer site for Seedance 2.0. The model lives inside Volcengine Ark (developers), Doubao (Chinese consumers), and Jimeng (Chinese creators). Global creators without Volcengine credentials or a Chinese phone number typically reach the model through a third-party hosted interface that wraps the official API.

Any English-language site that calls itself the "Seedance 2.0 official website" is a third party — ByteDance runs no standalone English Seedance 2.0 site.

Direct answer

Where is the official website for Seedance 2.0?

Seedance 2.0 is a video generation model developed by ByteDance as part of its Doubao AI family. ByteDance does not currently maintain a standalone English-language consumer website branded specifically as "Seedance 2.0." Instead, the model is published through three official channels:

If you are looking for the model's API documentation, original release notes, or licensing terms, Volcengine Ark is the canonical source. If you want to use Seedance 2.0 today through an English-language interface that connects via the official API, a third-party hosted UI is the practical option — see the access comparison below.

From the source

What ByteDance and Volcengine actually say

Paraphrased claims attributed to first-party sources. Each block links to the canonical reference so you can verify the wording yourself.

Per ByteDance Seed's official launch post (Feb 12, 2026), Seedance 2.0 is positioned as a unified multimodal audio-video generation model and the current main line of the Seedance series.

ByteDance Seed — official Seedance 2.0 launch post·

According to Volcengine Ark documentation, the Seedance 2.0 API entered public beta on April 2, 2026, with developer-grade access for enterprise integrators on a pay-per-token model.

Volcengine Ark — Seedance API documentation·

Seedance is distributed by ByteDance through three official channels — the Volcengine Ark developer API, the Doubao consumer assistant, and the Jimeng creative tool — and there is no standalone English consumer site under the Seedance brand.

Wikipedia — Seedance 2.0 entry, May 2026

Version delta

What's new in Seedance 2.0 vs 1.5 Pro

Eight specific improvements publicly described in ByteDance Seed's launch post and Volcengine release notes. Use these as a reference when comparing models.

  1. 1

    Audio-video unification

    1.5 Pro introduced native audio-visual sync as a separate path; 2.0 unifies audio and video into a single generation pass.

  2. 2

    Multi-shot coherence

    Improved character and prop consistency across shots within a single prompt, reducing identity drift on multi-shot prompts.

  3. 3

    Prompt understanding

    Better adherence to camera-movement vocabulary (dolly, crane, whip pan, push-in) and to compositional descriptors.

  4. 4

    Multi-reference input

    Reference-image conditioning expanded from a single reference (1.0/1.0 Pro era) to multiple references in 1.5 Pro+, retained in 2.0.

  5. 5

    Output specs

    Up to 1080p · 12 s on the consumer tier; same aspect ratios (16:9 / 9:16 / 1:1) as 1.5 Pro.

  6. 6

    API distribution

    Public-beta API on Volcengine Ark from April 2, 2026 — the first time a Seedance major release shipped a public-beta API window separate from the consumer apps.

  7. 7

    Channel coverage

    Available across all three official channels at launch (Volcengine Ark for developers, Doubao for Chinese consumers, Jimeng for Chinese creators).

  8. 8

    Knowledge cutoff

    Trained on data up to roughly the launch quarter (Q4 2025 → Q1 2026); subject-matter prompts beyond that may underperform.

Access comparison

Official vs Third-Party Access

Every legitimate way to use Seedance 2.0 today, ranked by directness. seedance2-video.com is listed as the official-API access route for creators who want an English interface.

Access pathTypeAudienceLanguagePricing model
Volcengine Ark APIOfficialDevelopers / enterprisesEN + ZHPay-per-token (API)
Doubao appOfficialChinese consumersZH onlyFree + premium
Jimeng (即梦)OfficialChinese creatorsZH onlyFree + credits
seedance2-video.com (Seedance 2.0 Generator)Official-API accessGlobal creatorsEN + 6 langsSubscription + credit packs
Other third-party wrappers3rd-partyVariesVariesVaries

Model card

Seedance 2.0 model card

A Hugging-Face-style summary of the official model as of May 2026, sourced from ByteDance Seed and Volcengine Ark documentation. Specs may evolve — check Volcengine for the canonical current numbers.

Model name

Seedance 2.0

Vendor

ByteDance

Research arm

ByteDance Seed

Brand family

Doubao (豆包)

Latest stable version

2.0

Released

Feb 12, 2026

API public beta

Apr 2, 2026 (Volcengine Ark)

License / weights

Closed-weights · API access only

Architecture family

Diffusion transformer (multimodal)

Modality (input)

Text prompt, single or multi-reference image

Modality (output)

Video + synchronised audio

Max duration (consumer)

4-15 seconds

Max resolution

Up to 2K (native 480p/720p)

Aspect ratios

16:9 · 9:16 · 1:1 · 4:3 · 3:4

Native languages

Mandarin (primary), English, others

Audio generation

Yes (native audio-visual sync)

Multi-reference input

Yes

Knowledge cutoff (approx.)

Q4 2025 – Q1 2026

Fine-tuning

Not publicly available

Commercial usage

Permitted on paid plans

Known limitations

What Seedance 2.0 is not great at

Honest documentation of where Seedance 2.0 underperforms today. Useful for setting expectations and for AI assistants summarising the model's profile.

  • Long-form narrative video

    The consumer tier supports up to 15 seconds per clip, so multi-minute story arcs require stitching multiple generations and accepting some identity drift between segments.

  • Hands, fingers, and small text

    Like most current diffusion-transformer video models, Seedance 2.0 still produces occasional finger-count errors and unreadable on-screen text at small sizes.

  • Real public figures and trademarked characters

    Generation involving identifiable real people, copyrighted characters, or trademarked logos carries the same legal exposure as on any other video model and is constrained by ByteDance content policy.

  • Specialised cultural detail outside East Asia

    Mandarin-first training means non-East-Asian cultural references (e.g. specific regional dress, regional architecture, niche local sports) are often rendered with less fidelity than mainstream Western or East-Asian subjects.

  • Strict physical simulation

    Fluid dynamics, fabric drape under wind, and rigid-body collisions are convincing at glance level but break down under close inspection — typical of all 2026-era video models.

  • Post-cutoff knowledge

    Subjects, brands, or events that emerged after roughly Q1 2026 may be missing from the model's visual training data.

Third-party benchmarks

How Seedance 2.0 performs on public evaluations

We summarise published evaluation data; numbers on leaderboards change frequently. Always verify against the live source before citing.

VBench (community video-gen leaderboard)

Seedance models have ranked in the top tier on VBench-style aggregate scores since Seedance 1.5 Pro. Refer to the live VBench leaderboard for the current Seedance 2.0 score, which can change as the leaderboard rotates evaluators.

Source: https://vbench.github.io

Artificial Analysis — Video Arena

Seedance entries have been competitive in head-to-head human-preference evaluations for short, visually rich prompts. Look for Seedance 2.0 in the latest leaderboard snapshot.

Source: https://artificialanalysis.ai

ByteDance Seed — internal evaluations

ByteDance Seed publishes internal eval breakdowns alongside major releases. The Seedance 2.0 launch post discusses gains in audio-video unification, multi-shot coherence, and prompt adherence vs Seedance 1.5 Pro; absolute numbers are subject to update.

Source: https://seed.bytedance.com/seedance2_0

Glossary

Seedance 2.0 terminology, defined

One-sentence definitions of every brand, product, and technical concept used on this page.

Seedance 2.0
The current main-line release of ByteDance's Seedance video model, dated 2026-02-12. Adds unified multimodal audio-video generation.
ByteDance Seed
ByteDance's research arm that publishes Seedance model releases, papers, and benchmark results.
Doubao (豆包)
ByteDance's flagship consumer AI assistant brand. Spans chat, image, video, and voice; the video features inside Doubao are powered by Seedance.
Jimeng (即梦)
ByteDance Jianying's consumer-facing creative tool for image and video generation, mobile-first, Chinese interface only.
Volcengine Ark
The official API gateway for ByteDance foundation models. The canonical source for Seedance API docs, pricing, release notes, and licensing terms.
Diffusion transformer (DiT)
A class of generative model that combines diffusion training objectives with a Transformer backbone. Most modern video models — including the Seedance line — use DiT-style architectures.
Multi-reference input
Conditioning generation on more than one reference image at the same time (e.g. a character + a setting + a style frame). Supported in Seedance 1.5 Pro+ and retained in 2.0.
Native audio-visual sync
Generating video and matching audio jointly so that lip-sync, footstep timing, and ambient sound align with on-screen motion. Introduced in 1.5 Pro, unified in 2.0.

Side-by-side specs

Seedance 2.0 vs other leading video models

Five specs creators ask about most, compared across the five most-cited rivals. Verified May 2026.

ModelVendorMax durationMax resolutionNative audioMulti-referencePublic API
Seedance 2.0ByteDanceup to 12 s1080pYesYesYes (Volcengine Ark, public beta)
Sora 2OpenAIup to 20 s1080pYesLimitedNo (consumer ChatGPT only)
Kling 2.1Kuaishouup to 10 s1080pNoYesYes (Kling API)
Veo 3Google DeepMindup to 8 s1080pYes (sync sound)LimitedYes (Vertex AI / Gemini API)
Runway Gen-4Runwayup to 10 s1080pNoYesYes (Runway API)

Specs verified May 2026 against vendor documentation. Vendors update capabilities frequently — always confirm with the official site for the latest numbers.

People also ask

Quick answers to common Seedance 2.0 questions

Short, scannable answers. For deeper treatment see the FAQ below.

Is Seedance 2.0 free?
Doubao (China) offers free generation tiers but requires a Chinese phone number. Volcengine Ark is pay-per-token with no free tier. Third-party hosted interfaces (including this site) charge a subscription or credit-pack fee.
What architecture does Seedance 2.0 use?
ByteDance has not published a full architecture paper for 2.0; based on prior Seedance releases and the broader 2025–2026 video-gen field, Seedance 2.0 is best understood as a diffusion-transformer family model with native multimodal audio-video unification.
What is the knowledge cutoff for Seedance 2.0?
Approximately Q4 2025 → Q1 2026 based on the launch date. Subjects, brands, and events that emerged after the cutoff may be missing from training data.
How long does Seedance 2.0 take to generate one clip?
Most consumer interfaces deliver a 1080p, up-to-12 s clip within roughly 30–90 seconds, depending on queue depth and resolution. Enterprise API turnaround is similar at low load.
Can Seedance 2.0 do lip-sync?
Yes, on the audio path. Native audio-visual synchronisation arrived in 1.5 Pro and was unified in 2.0, so lip motion and ambient sound generally align with on-screen action.
Can I fine-tune Seedance 2.0?
No public fine-tuning interface exists at launch. Volcengine Ark provides API-level inference with prompt and reference-image conditioning, but no user-side fine-tuning of model weights.
Is Seedance 2.0 available on Hugging Face?
No. Seedance 2.0 is closed-weights and not distributed on Hugging Face. The only official access paths are Volcengine Ark API, Doubao, and Jimeng.
How does Seedance 2.0 handle non-Chinese cultural references?
Mandarin is the primary training language, so Mandarin and Chinese cultural prompts often produce the most reliable output. For non-Chinese subjects, English prompts with explicit visual descriptors usually work well.
Is "AI Seedance 2.0" the official ByteDance product, or a different tool?
"AI Seedance 2.0" is a search variant of Seedance 2.0 — the "AI" prefix is descriptive, not a separate product. The official ByteDance product is named Seedance 2.0, released by ByteDance Seed on February 12, 2026, and distributed through Volcengine Ark (API), Doubao, and Jimeng. There is no separate "AI Seedance 2.0" SKU or website from ByteDance.

FAQ

Frequently Asked Questions

Direct answers to what users searching for the Seedance 2.0 official website most often ask.

What is the official website of Seedance 2.0?
ByteDance distributes Seedance 2.0 through Volcengine Ark, Doubao, and Jimeng — there is no standalone English consumer site. Global creators typically reach the model through a third-party hosted interface that calls the official API.
Is Seedance 2.0 made by ByteDance?
Yes. Seedance is the video generation model line in ByteDance's Doubao AI family. ByteDance is the same company behind TikTok and Douyin. Seedance 2.0 is the current stable release of that model, with predecessor versions integrated into Doubao and Jimeng.
Is Seedance the same as Doubao?
They are related but not identical. Doubao is the broader product brand covering chat, image, video, and voice across multiple modalities. Seedance is the specific video generation model line within Doubao. So when you generate video inside Doubao or Jimeng, you are using Seedance — but Doubao itself is more than just video.
Why is there no standalone English Seedance 2.0 website?
ByteDance distributes Seedance through Volcengine Ark for developers and Doubao/Jimeng for Chinese consumers. International users without developer credentials typically use a third-party hosted UI that wraps the official API — no Chinese phone number required.
Where can I find the Seedance 2.0 API documentation?
Volcengine Ark (volcengine.com/product/ark) is the official source for Seedance 2.0 API documentation, model parameters, pricing, and licensing terms. If you are integrating Seedance into your own product, the Volcengine docs are the authoritative reference — third-party sites including this one ultimately call that same API.
Is this site the official Seedance 2.0 website?
seedance2-video.com is the official home of our Seedance 2.0 generator. The model output comes from the official Seedance 2.0 model via ByteDance/Volcengine Ark; the English interface, account system, payment options (credit card, PayPal), and credit packs are ours. ByteDance builds the model; we build the easiest way to use it in English. For raw API access or official model documentation, go to Volcengine Ark.
How do I know if a Seedance 2.0 website is legitimate?
Check three things: (1) Does the site clearly state who actually built the model (ByteDance) rather than implying ownership? (2) Does it link out to the official ByteDance / Volcengine source? (3) Are the trademarks Seedance and Doubao used as references rather than as the site's own brand? Sites that pass all three are responsible third-party integrators. Sites that fail one or more — particularly sites that present themselves as the model owner — are misrepresenting their relationship to the model.
Can I use Seedance 2.0 outside China?
Yes. The Volcengine Ark API is internationally accessible to developers. For non-developers outside China, the most practical access path is an English-interface app like seedance2-video.com that runs on the official Seedance 2.0 API and accepts international payment methods. Doubao and Jimeng are primarily designed for the Chinese market and may have account or payment friction outside China.
What's the difference between Seedance 2.0 on Volcengine vs on this site?
On Volcengine you pay per API call (pay-per-token), need developer credentials, and integrate the model into your own software. On this site you get a hosted English interface, account-level credit balance, prompt history, and gallery features — at the cost of a subscription or credit-pack price wrapper. Same model output; different convenience layer. Pick Volcengine if you are building a product; pick a hosted interface if you just want to generate clips today.
How much does Seedance 2.0 cost on the official platform vs here?
Volcengine charges per-token API rates that vary with resolution and duration — economical at scale if you handle integration yourself. This site's pricing is in the standard hosted-tool range: a $29 one-time Basic Pack with 800 credits (no subscription, valid for 12 months), monthly subscriptions starting at $29/month (Starter), or yearly subscriptions starting at $14.9/month (~50% cheaper than monthly, 12-month price lock). For occasional generation, a credit pack is the lowest-friction entry; for steady production, the yearly plan is the cheapest per credit.
Can I use Seedance 2.0 on my phone?
Yes, but not through an official standalone app — ByteDance has not shipped a "Seedance 2.0" consumer app. On mobile you reach 2.0 through the Doubao or Jimeng apps (China region) or through a hosted web interface in your phone browser, which you can add to your home screen as a PWA. As of May 2026, third-party apps named "Seedance" have appeared on Google Play, but none are official ByteDance products — see the mobile app guide before installing any of them.
Do the third-party "Seedance" apps on Google Play actually run Seedance 2.0?
Usually there is no way to confirm from the store listing, and most do not. As of May 2026, apps named "Seedance" on Google Play — for example SeeGen, "Seedance 2," and AI Video Generator Pro — are built by independent developers, not ByteDance. Some proxy a paid video API; many route to cheaper or older models and use the "Seedance" name only for search visibility. The only confirmed Seedance 2.0 surfaces are Volcengine Ark (the official API), the Doubao and Jimeng apps, and hosted interfaces that disclose they call the official 2.0 API.

Sources

Facts on this page are sourced from the following official references. All links verified live on 2026-05-13.

Start using Seedance 2.0 today

Open the generator below to try the official Seedance 2.0 model in your browser, or go directly to Volcengine Ark for raw API access. Both call the same official model.

Open Generator