Contents
What is Seedance?Seedance version historySeedance versions comparedWhich Seedance version should you use?How to use Seedance in RekreateSeedance prompting tipsFrequently asked questionsSeedance is ByteDance's family of AI video generation models, developed by the ByteDance Seed team and offered through Dreamina/Jimeng, Doubao and the BytePlus ModelArk API. In about a year it went from silent 5–10-second clips (Seedance 1.0, June 2025) to joint audio-video generation (1.5 Pro), multimodal references across text, images, video and audio (2.0) and 30-second single-pass videos with up to 50 references (2.5, July 2026).
This guide explains each version, how they differ, and how to use Seedance 2.0 and 2.5 for short-form video in Rekreate.
- Latest: Seedance 2.5 (launched July 31, 2026) — up to 30 s per generation, up to 50 references, native audio, local editing.
- Most popular workhorse: Seedance 2.0 / 2.0 Fast (February 2026) — omni-reference with text, image, video and audio inputs.
- Audio arrived in Seedance 1.5 Pro (December 2025).
- In Rekreate: Seedance 2.0 Fast (4–15 s, 720p, up to 20 references) and Seedance 2.5 (4–30 s, 720p/1080p, up to 50 references).
What is Seedance?
Seedance is a text-to-video and image-to-video model family from ByteDance Seed, the research group behind the company's foundation models. Its defining traits are strong prompt adherence, smooth multi-shot storytelling and — since version 2.0 — a unified architecture that takes text, images, video clips and audio as references in one generation, producing video with synchronized sound.
For creators, that means you can hand Seedance a character photo, a product image, a reference clip for camera movement and a voice line, and get back a coherent scene instead of stitching separate tools together. It is one of the most used models for UGC ads, product shots and cinematic short-form content in 2026.
Seedance 2.0 also attracted controversy: shortly after launch, the Motion Picture Association and several studios accused ByteDance of generating copyrighted characters, and Disney sent a cease-and-desist letter. Responsible platforms add safeguards — for example, Rekreate checks uploads for real people and rejects prohibited inputs.
Seedance version history
Seedance 1.0 (Lite and Pro)
ByteDance's first public video model. Pro produced up to 1080p clips of 5–10 seconds with native multi-shot storytelling; Lite traded quality for speed and cost (commonly 720p). Silent video only.
Seedance 1.0 Pro Fast
A faster, cheaper variant of 1.0 Pro aimed at high-volume API usage and iteration.
Seedance 1.5 Pro
The first Seedance with joint audio-video generation: dialogue, sound effects and ambience created in the same pass, multilingual lip-sync, cinematic camera control and 4–12-second clips up to 1080p.
Seedance 2.0 In Rekreate
A unified multimodal audio-video architecture accepting text, image, video and audio inputs together. Adds omni-reference (reuse a character, product, motion or sound from references), first/last-frame control and video editing, with generations up to 15 seconds.
Seedance 2.0 Fast In Rekreate
Lower-latency, lower-cost tier of 2.0 — ideal for testing prompts and producing many variations. In Rekreate: 4–15 s at 720p with up to 20 references.
Seedance 2.5 In Rekreate
30-second single-pass generation, up to 30 images, 10 video clips and 10 audio clips as references (50 total), timestamp-level local editing, stronger shot transitions and multi-round extension toward multi-minute stories.
Dates from ByteDance Seed announcements and press coverage. Resolution tiers differ between ByteDance's own apps and third-party APIs.
Seedance versions compared
| Capability | 1.0 Pro | 1.5 Pro | 2.0 | 2.5 |
|---|---|---|---|---|
| Max length per generation | 10 s | 12 s | 15 s | 30 s |
| Native audio | ✕ | ✓ | ✓ | ✓ (improved long-form consistency) |
| Input types | Text, image | Text, image | Text, image, video, audio | Text, image, video, audio |
| Max references | 1 image | Image(s) | Multiple (up to 20 in Rekreate's Fast tier) | Up to 50 (30 img + 10 video + 10 audio) |
| Editing | ✕ | ✕ | Reference-based editing | Timestamp-level local editing |
| Available in Rekreate | ✕ | ✕ | ✓ (2.0 Fast) | ✓ |
Which Seedance version should you use?
- Quick tests, many variations, 15 seconds or less: Seedance 2.0 Fast. It is the cheapest way to explore hooks and camera ideas.
- Final ads, longer stories, many references: Seedance 2.5. Use it when you need 20–30 seconds in one take, 1080p, or a product plus multiple characters and reference motions.
- First and last frame control: Seedance 2.0's first/last-frame mode, useful for transitions and before/after reveals.
- Legacy workflows: Seedance 1.x models are still offered by some APIs but are superseded by 2.x in quality, audio and control.
For talking avatars with long scripts, an audio-driven avatar model like OmniHuman may be more consistent; for motion copied exactly from a video, use Kling Motion Control.
How to use Seedance in Rekreate
Open Seedance and pick the model
Choose Seedance 2.0 Fast for drafts or Seedance 2.5 for final renders.
Add references
Upload character and product images, a reference video for motion or framing, and optional audio. Tag them in the prompt (for example @image_1 holds @image_2).
Write a shot-based prompt
Describe subject, action, camera, lighting and sound per shot. Keep the first two seconds visually striking — that's your hook.
Set duration and resolution
Pick 4–15 s (2.0 Fast) or 4–30 s and 720p/1080p (2.5), then generate. Results appear in your history and can be shared to the community gallery.
Seedance prompting tips
- Structure by shots. "Shot 1 (0–3 s): close-up… Shot 2 (3–8 s): wide…" gives Seedance clear edit points.
- Name camera moves. Dolly in, handheld, orbit, whip pan, top-down — Seedance follows camera language well.
- Describe sound explicitly. Dialogue in quotes, plus ambience and SFX ("quiet café murmur, cup clink").
- Use references for identity, prompts for action. Let images lock faces and products; use text for what they do.
- Avoid real people and copyrighted characters. Use original characters or licensed brand assets.
Frequently asked questions
What is Seedance AI?
Seedance is ByteDance's AI video generation model family. It creates videos from text, images, video and audio references, with native synchronized audio since version 1.5 Pro and 30-second generations in version 2.5.
What is the latest version of Seedance?
As of September 2026 the latest version is Seedance 2.5, announced on June 23, 2026 and launched on July 31, 2026.
What is the difference between Seedance 2.0 and 2.5?
Seedance 2.5 doubles single-pass length from 15 to 30 seconds, raises references to up to 50 (30 images, 10 videos, 10 audio clips), adds timestamp-level local editing and improves shot transitions and audio consistency over longer videos.
Does Seedance generate audio?
Yes. Seedance 1.5 Pro introduced joint audio-video generation, and Seedance 2.0 and 2.5 generate dialogue, sound effects and ambience in sync with the video.
Is Seedance free?
ByteDance's consumer apps offer limited free usage in some regions, and platforms like Rekreate offer free trials. Heavy or commercial use is paid via credits or API pricing.
Can I use Seedance outside China?
Yes. Seedance is available internationally through BytePlus ModelArk and third-party platforms, including Rekreate, which offers Seedance 2.0 Fast and Seedance 2.5.
Seedance vs Kling vs Veo — which is best?
Seedance leads on multimodal references and long single-pass clips, Kling 3.0 on multi-shot control and motion transfer, and Veo 3.1 on cinematic realism and audio. See our Seedance vs Kling vs Veo comparison.
Create your next viral video with Rekreate
Find a proven short-form video, recreate it with your script, avatar and voice, and generate scenes with Seedance, Kling and Veo in one workspace.
Start free trial →