August 07, 2026

Seedance 2.5 is now available.

ByteDance's next-generation video model is live on LetzAI: up to 30-second native clips, up to 50 multimodal references, native audio, and 480p/720p at launch.

More Realism, Longer Takes, Deeper Reference Control

Seedance 2.5 is now available on LetzAI for all users. It is our latest video generation model from ByteDance, built for stronger production readiness: up to 30 seconds in a single native render, up to 50 multimodal references per generation, better audio-visual alignment, and smoother long-take workflows than Seedance 2.0.

Our own benchmarks have shown that the model performs especially well on realistic content, such as product walkthroughs, brand vignettes, and videos featuring real people (both cinematic and handheld social media clips).

Seedance 2.5 side by side with other top AI video models

For now, Seedance 2.5 runs at 480p and 720p. Higher resolutions are coming later. Native audio is included on every generation.

Up to 50 Multimodal References

This is the biggest jump in creative control. Seedance 2.5 accepts up to 50 reference files in a single pass, far beyond Seedance 2.0's 12-file budget. Official limits are split across three budgets (they are not interchangeable):

Up to 30 images: characters, products, environments, style boards, wardrobe, composition
Up to 10 video clips: motion templates, camera moves, pacing guides
Up to 10 audio files: music, dialogue, ambience to sync to

On LetzAI, that runs through the same Omni Reference workflow as Seedance 2.0, now scaled to the full 2.5 budget. Upload your files as cards below the prompt, then address them in natural language. Feed a full reference library to lock faces, products, environments, and motion across an entire 30-second take.

On the LetzAI Canvas, you can reference each element on the board directly when generating with Seedance 2.5: characters, products, frames, and other assets already laid out in your scene. If you want to reuse them later, save them to memory so the agent can pull the same references into future generations without re-uploading or hunting them down again.

For simpler workflows, switch to First and Last Frame mode to lock start and end keyframes.

References of real people

Just like Seedance 2.0, references of real people and IPs are limited. In order to unlock image rights, you need to be on an Enterprise plan and go through our copyright clearance process.
Enterprise users see a "Seedance 2.5. Enterprise" version, with the same capabilities as Seedance 2.5, but with the extended ability to reference real people and IPs, and to create mature content.

Click here to learn more about Enterprise plans

Up to 30 Seconds, One Continuous Clip

Seedance 2.5 generates 4 to 30 seconds per take, double the previous 15-second cap. That is one continuous clip with pacing, scene transitions, and spatial continuity baked in. No stitching short generations together, and no seams at the cut points.

That shape matches what creators actually ship: social ads, product walkthroughs, brand vignettes, and short narrative beats. A 30-second take can carry a hook, a turn, and a payoff in one generation, with enough reference budget to keep every subject consistent from first frame to last.

Native Audio, Always On

Every Seedance 2.5 generation includes audio at no extra cost. There is no toggle and no surcharge. Audio and picture are co-processed in the same pass, with stronger audio-visual alignment than previous Seedance generations. That is especially useful when you upload music or dialogue as one of your audio references.

Director-Level Camera Control

Through your prompt or video references, you can still direct cinematic camera work: dolly zooms, tracking shots, handheld drift, POV switches, and multi-axis moves. With up to 30 seconds of native output and up to 10 video references, there is room for a full camera arc, not just a single move in isolation.

Technical Specs

Duration: 4 to 30 seconds
Resolution at launch: 480p and 720p (higher resolutions coming later)
Aspect ratios: auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
Modes: Text-to-Video, Image-to-Video, Reference-to-Video, First + Last Frame
Audio: always included
References: up to 50 files (30 images / 10 videos / 10 audio)

Pricing

Seedance 2.5 is priced per second of generated video. Audio is always included.

480p: 150 credits per second (~€0.18 / ~$0.20)
720p: 300 credits per second (~€0.35 / ~$0.40)

EUR/USD shown at our best-volume Enterprise Large top-up rate. Actual euro cost depends on your plan.

These prices are among the cheapest in the industry for any model of this quality, served on aggregator platforms like LetzAI.

For example, a 30-second clip at 720p costs 9,000 credits; at 480p it costs 4,500 credits (a good option for drafts). A 10-second 720p draft costs 3,000 credits. Full credit tables for every model are on letz.ai/pricing (Credit Costs).

See our Enterprise Plans for Enterprise setups.

Getting Started

Open the Animate tab on letz.ai
Select Seedance 2.5 in the model selector
Choose Omni Reference or First and Last Frame
Write your prompt, upload references (up to 30 images / 10 videos / 10 audio), set duration (up to 30s), and click Animate

Seedance 2.0 Is Still Here

Seedance 2.5 does not replace the rest of the lineup. Seedance 2.0 and Seedance 2.0 Enterprise remain available for cinematic clips up to 15 seconds with a 12-reference budget (including higher resolutions where supported). Use 2.0 for shorter shots; pick 2.5 when you need a longer native take and deeper reference control.