Seedance 2.5: 30-Second Native 4K Video, 50 References, Coming July 2026
ByteDance unveiled Seedance 2.5 at the Volcano Engine FORCE Conference. It generates native 30-second 4K video in a single pass, accepts up to 50 multimodal reference inputs, and is expected to launch in early July 2026.
On June 23, 2026, ByteDance unveiled Seedance 2.5 at the Volcano Engine FORCE Conference. The model is currently in global enterprise beta testing and is expected to launch publicly in early July 2026. Here is a full breakdown of what is new and how it compares to Seedance 2.0.
The headline upgrades
- Native 30-second single-segment video — no stitching required. Seedance 2.0 maxes out at 15 seconds.
- Up to 50 multimodal reference inputs simultaneously — images, audio clips, style references, and 3D white models. Seedance 2.0 accepts a maximum of 12.
- Native 4K output with 10-bit color depth. Seedance 2.0 tops out at 1080p.
- Approximately 20% improvement in prompt adherence.
- Controllable video editing with localized scene modification.
- 3D previz support for pre-production workflows.
Seedance 2.5 vs 2.0 at a glance
| Seedance 2.0 | Seedance 2.5 | |
|---|---|---|
| Max resolution | 1080p | 4K (10-bit) |
| Max duration | 15 seconds | 30 seconds (native) |
| Reference inputs | Up to 12 (images + video + audio) | Up to 50 (full multimodal) |
| 3D previz | No | Yes |
| Scene editing | No | Localized / controllable |
| Prompt adherence | Baseline | +20% improvement |
| Launch status | Available now | Early July 2026 |
Why 30 seconds natively matters
The shift from 15 to 30 seconds in a single pass is more significant than the number suggests. Multi-segment stitching — generating two clips and joining them — introduces visible discontinuities in motion, lighting and camera direction. Native 30-second generation maintains a single coherent trajectory throughout, which matters most for narrative sequences, music video cuts and product demos that need a continuous arc.
What 50 reference inputs unlocks
Seedance 2.0 already led the market on reference handling with up to 12 inputs. At 50, the model moves into territory that was previously only possible in professional compositing pipelines: feeding in a complete product line, a full wardrobe, multiple location references and an audio track simultaneously. For commercial production, that means fewer generation rounds and tighter brand consistency across a full campaign.
Already building with Seedance 2.0?
Seedance 2.0 is available on FAV now across all plans — 480p through 4K, up to 15 seconds, with 12 reference inputs. When Seedance 2.5 launches, it will be available on the same account and credit balance.
When and where to access it
ByteDance announced an early July 2026 public launch window. As of this writing the model is in global enterprise beta. We will add Seedance 2.5 to FAV as soon as API access is available — existing users will not need to create a new account or purchase new credits.
Try Seedance 2.0 today while you wait for 2.5.
Available on every plan. Free credits on sign-up. Seedance 2.5 coming July 2026.
Frequently asked questions
When does Seedance 2.5 come out?
ByteDance announced an early July 2026 launch. The model was unveiled on June 23, 2026 at the Volcano Engine FORCE Conference and is currently in global enterprise beta testing.
What is the maximum video length in Seedance 2.5?
Seedance 2.5 generates a native single-segment video up to 30 seconds. This is double the 15-second maximum of Seedance 2.0 and requires no clip stitching.
Does Seedance 2.5 support 4K?
Yes. Seedance 2.5 outputs native 4K with 10-bit color depth. Seedance 2.0 supports up to 1080p.
How many reference inputs does Seedance 2.5 accept?
Up to 50 multimodal references simultaneously, including images, audio clips, style references and 3D white models. Seedance 2.0 accepts up to 12.
Where can I use Seedance 2.5?
Seedance 2.5 will be available on FAV (foraivideo.com) when the public API launches in early July 2026. Seedance 2.0 is available now on every plan.
What is the difference between Seedance 2.0 and Seedance 2.5?
The main upgrades in 2.5 are native 30-second video (vs 15s), 50 reference inputs (vs 12), native 4K with 10-bit color (vs 1080p), and controllable scene editing. Both share the same core multi-reference architecture.
Models in this article
Multi-modal reference: up to 9 images, 3 videos, 3 audio tracks in one generation.
Same multi-reference power at lower cost and faster turnaround. Max 720p.
Keep reading
How Seedance 2.5 Could Change the Way Short-Form Video Gets Made
Every generation of AI video models has changed what a small team can produce alone. Seedance 2.5 — native 30-second 4K, 50 references, controllable editing, 3D previz — is positioned to shift short-form production further: from single clips to fuller scenes, and from one-shot gambles toward an editable pipeline.
Read article →How Long Can Seedance 2.5 Videos Be? 30-Second Native Generation Explained
Seedance 2.5 generates up to 30 seconds of continuous video in a single pass — no stitching, no seams. Here is exactly what that means, why "native" duration is the important word, and how to plan a shot that uses the full length well.
Read article →Seedance 2.5 Benefits: 6 Reasons It Is the Biggest AI Video Upgrade of 2026
Seedance 2.5 is not an incremental update. Native 30-second 4K video, 50 reference inputs and controllable scene editing move it from "AI video generator" toward "production tool". Here is what each upgrade actually changes.
Read article →