What Is Seedance 2.0? ByteDance's AI Video Model Explained (Plus What 2.5 Changed)
May 23, 2026 · Updated September 30, 2026 · 7 min read
Seedance 2.0 is ByteDance's AI video model: from a text prompt, optionally guided by up to 12 reference files (at most 9 images, 3 video clips and 3 audio clips), it generates a 4–15-second clip with sound. Released in February 2026, it's no longer the newest Seedance — Seedance 2.5 arrived on July 31, 2026 with 30-second generations — but 2.0 is still widely offered and ranks 5th in Artificial Analysis's blind-vote text-to-video arena (with audio) as of September 30, 2026.
Seedance 2.0 vs 2.5 at a glance
| Seedance 2.0 | Seedance 2.5 | |
|---|---|---|
| Released | February 2026 | July 31, 2026 |
| Longest single generation | 15 s (4–15 s) | 30 s, plus further extension rounds |
| Reference inputs | Up to 12 files: at most 9 images, 3 videos, 3 audio clips | Up to 30 images, 10 videos, 10 audio clips |
| Resolution | 480p–1080p, plus native 4K on BytePlus; Fast and Mini variants max 720p | Not stated in ByteDance's launch post; varies by platform |
| Official API price | $0.15/s at 720p, $0.37/s at 1080p (BytePlus) | Not in the announcement; varies by host |
| Arena rank, with audio | #5 (Elo 1,210) | Not ranked yet |
| On AI Growth Kit | Yes | No |
Sources: ByteDance Seed and CapCut announcements, BytePlus's published rates (July 2026) and the Artificial Analysis leaderboard, checked September 30, 2026.
What Seedance 2.0 does
Seedance 2.0 generates picture and sound in one pass: sound effects, ambience and music come out with the clip instead of being added afterwards. It accepts four kinds of input — text, images, video and audio. Images can pin down a character or a look, a video clip can show the camera move or motion you want, and audio can set the rhythm. Clips run 4–15 seconds in six aspect ratios.
It holds up well in blind votes. On Artificial Analysis's text-to-video arena with audio it ranks 5th, behind Gemini Omni Flash, Wan 3.0 and two versions of MiniMax H3, and ahead of Kling 3.0 (12th) and Veo 3.1 (14th). It is also 5th on the image-to-video board with audio (Elo 1,176).
On BytePlus, ByteDance's cloud platform, it's billed per second of output: $0.07 at 480p, $0.15 at 720p, $0.37 at 1080p and $0.78 at native 4K. The Fast variant costs $0.06–$0.12 and Mini $0.04–$0.08, both capped at 720p. A 15-second 1080p clip therefore costs about $5.55; the same clip in 4K is $11.70.
Where it falls short
- 15 seconds per generation. Longer scenes mean stitching clips or extending them — the main thing 2.5 fixes.
- Complex motion. Even for 2.5, ByteDance says the physical plausibility of complex motion and the stability of scenes where several subjects interact still need work.
- Uneven availability. After Hollywood studios accused the model of copyright infringement in February 2026, ByteDance paused its global rollout in mid-March. CapCut then launched it country by country from March 26, reaching the US on April 7 for paid users.
- Quality costs extra. The cheap Fast and Mini variants stop at 720p, and 1080p and 4K cost roughly 2.5× and 5× the 720p rate.
What Seedance 2.5 changed — and what the predictions missed
When we previewed 2.5 in June, the expectation was higher resolution, steadier motion, faster generation and better prompt following. Here is what ByteDance actually shipped on July 31, 2026:
- Length, not resolution, was the headline. 2.5 generates up to 30 seconds per pass and can be extended in further rounds. ByteDance's launch post doesn't state a resolution, and platforms differ on what they offer; a native 4K tier arrived on Seedance 2.0, through BytePlus, instead.
- Far more reference material. Up to 30 images, 10 video clips and 10 audio clips per generation, versus 2.0's cap of 12 files.
- Edits instead of re-rolls. Timestamp-level editing lets you change one stretch of a clip's picture or sound without regenerating the rest, and green-screen editing is supported.
- Consistency, with caveats. ByteDance says characters and settings stay consistent across extensions, but lists complex motion and multi-subject scenes as open problems.
- Speed is unconfirmed. ByteDance published no generation-time figures, and 2.5 isn't on Artificial Analysis's leaderboard yet, so there's no independent quality comparison either.
Where you can use Seedance (September 2026)
- ByteDance's Chinese apps: Jimeng AI and Doubao carry both 2.0 and 2.5.
- CapCut and Dreamina: 2.0 for paid users, including in the US since April 7; 2.5 for subscribers aged 16 and over, rolling out from July 31 in Europe, Asia, the Middle East and South America.
- APIs: BytePlus ModelArk is the official route for 2.0, at the prices above. ByteDance announced 2.5 API access as “coming soon” at launch; hosts such as Replicate already run it, with 4–30-second clips.
- Creative platforms: Runway includes Seedance 2.0 on its paid plans and 2.5 from its $35/month Pro plan; Pika charges 122 credits for a 5-second 720p Seedance 2.5 clip.
- AI Growth Kit: Seedance 2.0 — details below.
How to prompt Seedance 2.0
- Describe the shot, not just the subject. “A cup of coffee” is weak. “Slow push-in on a steaming cup of coffee, warm morning light, shallow depth of field” gives the model a camera move, light and focus.
- Name the camera move and the lighting. Pan, tilt, push-in, drone shot or static; golden hour, overcast or neon-lit. These words steer the whole frame.
- One idea per clip. Three actions in one prompt tend to blur together; generate separate clips and cut them together.
- Say what should be heard. Sound is generated with the picture, so describe it — rain on a window, crowd noise, no music.
- Choose the aspect ratio first. 9:16 for TikTok, Reels and Shorts; 16:9 for YouTube and landing pages; 1:1 for feed ads. Cropping later throws away resolution.
- Keep a prompt library. Save prompts that produced a usable camera move or look, and reuse them as templates.
More examples in how to write AI video prompts.
Running Seedance 2.0 on AI Growth Kit
Seedance 2.0 is the default model in our Video Generator. You write a prompt and choose 4–15 seconds, an aspect ratio (16:9, 9:16, 1:1, 4:3 or 3:4) and 720p or 1080p. It costs a flat 10 credits per second — 50 credits for a 5-second clip — at either resolution, so shorter drafts save credits and lower resolution doesn't. Seedance 2.0 also powers Video Extend (10 credits per second) and Video Edit (180 credits per edit, since an edit always returns a 15-second clip).
Credits cost $9 for 200, $29 for 750 or $59 for 1,800 as one-time packs that never expire, or $19 for 600 and $49 for 2,000 a month — roughly $0.25–$0.45 per second of video (see pricing). The 20 free credits you get at sign-up don't cover a clip; the shortest Seedance clip, 4 seconds, costs 40.
What we don't offer: Seedance 2.5, 4K, image or reference uploads (the form is text-to-video only), or audio settings. If you're a developer who only needs Seedance at 720p, BytePlus's $0.15 per second is cheaper than our flat rate; at 1080p the two are comparable ($0.37 against our $0.25–$0.45). Ours makes most sense when you also want Kling 3.0, Veo 3.1 and the other models on one balance — see how they compare in the best AI video generators in 2026.
Use it responsibly
Use AI video to visualize ideas, scenes and products — not to pass off invented events as real or to recreate protected characters and films. ByteDance added filters against unauthorized IP after the studio complaints, and clips made in CapCut and Dreamina carry invisible watermarks, visible AI labels and C2PA Content Credentials.