Treating “Seedance 2” as one interchangeable button is how teams burn credits on the wrong resolution — or ship a draft-looking clip as if it were a finish pass.
On February 12, 2026, ByteDance Seed announced Seedance 2.0 as a unified multimodal video model: text, image, audio, and video references into one generation path, with a strong story around motion stability and director-level control (Seed · Seedance 2.0 product page).
On SupaImagine, that family shows up as three generator lanes with separate money pages:
| Lane | Money page | Catalog role (site product) |
|---|---|---|
| Seedance 2 (full) | /video/seedance-2/ | Top fidelity: 480p–4K, highest quality grade |
| Seedance 2 Fast | /video/seedance-2-fast/ | Middle: 480p–720p, near-final polish, faster path |
| Seedance 2 Mini | /video/seedance-2-mini/ | Draft volume: 480p–720p, lowest-cost exploration |
This post is a decision page, not a host price sheet: what each lane is for, what official sources claim about Seedance 2.0, how SupaImagine splits the grades, and how to A/B them with the same prompts — including real motion samples from SupaImagine’s Explore showcase.
TLDR
Seedance 2.0 is ByteDance Seed’s multimodal video stack: joint audio-visual generation from text / image / audio / video inputs, positioned for stable motion and controllable camera language (official launch, product page). On SupaImagine, Mini is the draft volume lane (cheapest grade, 480p–720p), Fast is the near-final middle (same resolution ceiling as Mini, higher quality grade), and full Seedance 2 is the finish lane — the only of the three that opens 1080p and native 4K. Use Mini to explore, Fast when a draft almost ships, full when resolution and final polish matter.
Key Takeaways
- Official identity: Seedance 2.0 launched Feb 12, 2026 as a multimodal audio-video model with text/image/audio/video inputs (Seed blog).
- Product spine: Seed’s product page stresses motion stability, director-level control, and multi-reference workflows (seedance2_0).
- SupaImagine three-lane split (product fact, not three ByteDance marketing SKUs with identical public names): catalog maps full / Fast / Mini to different quality/speed grades and different resolution ceilings — only full exposes 1080p + 4K.
- Shared clip envelope on SupaImagine: all three lanes use the same default shape in catalog (16:9, default 720p, default 5s) with product pages describing 4–15s clip jobs, always-on synced audio framing, and the same input modes: text-to-video, image-to-video (first / first+last frame), and multimodal reference-to-video.
- Mode tip: first/last-frame image-to-video and multimodal reference-to-video are mutually exclusive on a single run — pick one path per task rather than stacking every reference type at once.
- How to choose: Mini → explore, Fast → near-final under full cost, full → ship at 1080p/4K. Do not write Mini drafts as if they were 4K finish files.
What Seedance 2.0 “actually shipped” (upstream)
| Claim | Source |
|---|---|
| Public launch Feb 12, 2026 | Seed official launch blog |
| Multimodal joint generation (text / image / audio / video in) | Same launch + product page |
| Motion stability + director-level control narrative | Seedance 2.0 product page |
| Industry coverage of real-world physics / joint generation framing | Forbes · Feb 12, 2026 |
| Three live generator lanes on SupaImagine | /video/seedance-2/, /video/seedance-2-fast/, /video/seedance-2-mini/ |
Honesty boundary: ByteDance’s public story is about Seedance 2.0 as a model family. SupaImagine’s Fast and Mini names are on-site grade lanes. Treat them as draft / mid / finish grades of the same product family, not as three independently marketed consumer SKUs with separate press releases.
Head-to-head on SupaImagine
Capability comparison (site product + catalog)
| Dimension | Mini | Fast | Full Seedance 2 |
|---|---|---|---|
| Money page | /video/seedance-2-mini/ | /video/seedance-2-fast/ | /video/seedance-2/ |
| Job | Low-cost draft volume | Near-final clips | Finish pass / hero export |
| Resolution ceiling (catalog) | 480p · 720p | 480p · 720p | 480p · 720p · 1080p · 4K |
| Quality / speed grade (catalog) | quality 3 · speed 4 | quality 4 · speed 4 | quality 5 · speed 2 |
| Inputs (all three) | text-to-video · image-to-video · multimodal reference-to-video | same | same |
| Default shape (catalog) | 16:9 · 720p · 5s | same | same |
| When to pick | Many tries, composition hunting | Draft almost ships, still ≤720p | Need 1080p/4K or max polish |
On the generators, image-to-video (first frame), first+last frames, and multimodal reference-to-video are mutually exclusive modes — pick one path per task rather than stacking every reference type at once. Confirm the live mode controls on each money page before you queue a long batch.
Scenario winners
1. Social drafts, storyboard previews, “which camera move?” → Mini
You still need five compositions to know the story. Mini is the volume lane: same multimodal inputs, lower quality grade, 480p–720p only. Iterate prompts here before you spend a finish pass.
2. Client-ready 720p ads, product loops, near-final social → Fast
When motion and audio already feel close, Fast is the middle grade: still capped at 720p, but catalog quality sits above Mini. Use it when you would be embarrassed to send a Mini draft but do not need 4K yet.
3. Hero cutdowns, launch films, anything that must survive a big screen crop → full Seedance 2
Full is the only lane with 1080p and 4K in catalog. Product-page spine on SupaImagine: finish-tier fidelity, synced audio, cinematic camera control, text / frame / multimodal refs. Promote winners from Mini/Fast here — do not invent a new prompt family on the expensive pass.
See the motion language (Explore samples)
These clips come from SupaImagine’s Explore / showcase Seedance video set. They illustrate what Seedance-style motion can look like — camera drive, commercial product rhythm, ink-style performance — not a claim that each file was rendered on Mini vs Fast vs full in a controlled A/B. Media is served from the site CDN (cdn.supaimagine.ai).
Sample A — one-shot station rush (camera energy)
Prompt shape to re-run (short form):
High-energy cinematic one-shot inside a massive Egyptian train station at dawn,
camera rushing through steam between passengers and luggage carts, jumping onto
a moving train beside windows, handheld shake, aggressive forward momentum,
fast tilt-up to the steel ceiling, warm sunlight piercing smoke, hyper-real
textures, 5–8 seconds, native synced ambience.Sample B — product commercial energy (finish-grade intent)
Use this job type (product hero + rhythmic edit energy) as a finish-lane test: draft composition on Mini or Fast at 720p, then re-run keepers on full Seedance 2 at 1080p/4K.
Sample C — ink / performance motion (stylized control)
Prompt shape (short form):
15-second cinematic clip in ink wash + watercolor style. A graceful dancer
moves slowly in empty white space; every motion leaves black ink and soft
pigment trails that spread like living brushstrokes. Minimal set, high contrast,
synced ambient score, no dialogue.Browse more Seedance-family results on the site’s Explore prompt library when you want additional motion references before you open a generator.
What We Know vs. What We Don’t
What we know
- Seedance 2.0 is a public ByteDance Seed multimodal video model with a dated launch post and product page (launch, product).
- SupaImagine exposes three live money pages with distinct catalog grades and resolution ceilings (full only for 1080p/4K).
- All three SupaImagine lanes share t2v / i2v / multimodal reference capability flags in catalog and the same default aspect/duration defaults.
- Explore already hosts a Seedance showcase video set with playable mp4 + poster assets on the site CDN (samples embedded above).
What we don’t know (or must qualify)
- Whether every third-party surface’s “Seedance 2 Fast / Mini” naming matches ByteDance’s own consumer product labels 1:1 — on SupaImagine, treat Fast/Mini as grade lanes.
- Independent third-party leaderboards that score Mini vs Fast vs full on an open prompt suite with fixed seeds.
- Whether a specific Explore clip was produced on Mini, Fast, or full — showcase metadata tags the family as Seedance, not the on-site grade used for that historical render.
- Exact credit math in this article — open each generator for live pricing on SupaImagine.
How to Evaluate on SupaImagine
Run the same prompt kit on Mini → Fast → full. Judge on: motion continuity, audio sync feel, prompt adherence, and whether the file survives a phone crop vs a large export.
Prompt kit (copy-paste)
1. Camera stress (favor full if the move must stay clean)
One continuous handheld push through a rainy neon alley at night, puddle
reflections, soft bokeh signs in Japanese and English, person with umbrella
walks toward camera, slow dolly-in, cinematic 5s, synced rain ambience.2. Product tabletop (finish on full at 1080p/4K when shipping)
Macro tabletop of a matte black wireless earbud case opening slowly, soft
studio light, subtle rotation, clean white seamless background, 5s product
hero, quiet whoosh synced to the lid.3. Character performance (draft on Mini first)
Full-body dancer in a sunlit warehouse, sharp footwork, camera orbits 90
degrees, natural skin texture, no morphing limbs, 5s, light room tone + soft steps.Three clicks
- Open Seedance 2 Mini — burn the kit at 720p.
- Open Seedance 2 Fast — re-run keepers.
- Open Seedance 2 — promote the winner at 1080p or 4K when the job is a ship file.
Decision rule (one screen)
| Your situation | Start here | Then |
|---|---|---|
| You need many directions in minutes | Mini | Promote 1–2 winners to Fast or full |
| Clip must feel almost final at ≤720p | Fast | Full only if you need 1080p/4K |
| Export is a hero / launch / big-screen cut | Full | Lock prompt on Mini first if still exploring |
| You only have one serious clip today | Full | Run the real deliverable prompt first |
| Input is frames or multimodal refs | Any lane | Respect mutually exclusive modes (first/last vs multimodal ref) |
FAQ
Are Mini, Fast, and full three different ByteDance consumer product names?
Treat SupaImagine’s Mini / Fast / full as on-site grade lanes of the Seedance 2 family. ByteDance’s public launch materials describe Seedance 2.0 as the multimodal model (launch post).
Which lane supports 4K?
On SupaImagine catalog, only full Seedance 2 includes 4K (and 1080p). Mini and Fast stop at 720p.
Do all three support image-to-video and reference video?
Yes in catalog: t2v, i2v, and multimodal v2v / reference style paths are declared for all three slugs. On the generator, first/last-frame i2v and multimodal reference modes are not stacked in one task.
How long can a clip be?
SupaImagine product pages for the family describe 4–15s clip jobs; catalog defaults use 5s. Confirm the live duration control on the generator you open.
Should I draft on full to “save time”?
Usually no. Full is slower (catalog speed grade 2 vs 4) and is the only 4K path — wasteful for composition hunting. Mini or Fast first, full for keepers.
Can I see more Seedance examples before generating?
Yes — use the embedded Explore samples above, then browse the Explore library for additional Seedance-family prompts and results.
Does Seedance 2 replace Veo or other video models on SupaImagine?
No. Seedance is one family; SupaImagine also offers other video models (for example Veo lanes). This article only decides which Seedance grade to open.
Is native audio always on?
SupaImagine’s Seedance family pages market synced / always-on audio as part of the product framing. Still verify the live controls on each generator before promising a silent export to a client.
Where should I start right now?
Open Seedance 2 Mini, run Prompt 1 from the kit, then re-run the better take on full Seedance 2 if you need a shippable resolution.
Will this article stay accurate if catalog rates change?
Resolution ceilings and grade roles are the durable product facts. Live credit rates can move — always trust the on-page generator, not a frozen blog table.
About Mara Ellison
Mara Ellison is an AI Image Model Analyst. She tracks model releases and product-lane splits so builders don’t have to, and writes the numbers-first breakdowns on SupaImagine — including video-family decision pages like this one.
