30-second storytelling
Build a complete 30-second segment, then extend it twice when the story needs more room.

Plan 30-second stories, coordinate multimodal references, interpret cinematic direction, and refine professional creative ideas with ByteDance's next-generation audio-video model.
Seedance 2.5 has been released by ByteDance. Nanobana integration is in progress, so generation is not available on this page yet.
Explore what is comingBuild a complete 30-second segment, then extend it twice when the story needs more room.
Coordinate combined image, video, and audio materials inside one richer creative brief.
Use a reference for framing, camera rhythm, and intent instead of relying on simple motion transfer.
Work toward white-model control, green-screen replacement, blocking, and broader audio-video changes.
Key Features
These four points summarize the model's published capabilities. They describe Seedance 2.5 itself, not controls currently exposed by Nanobana.
Feature 01 · 30-second narrative
Seedance 2.5 can create a single video segment up to 30 seconds and extend it twice. That gives setup, action, camera movement, and resolution more space before separate clips need to be assembled.

Feature 02 · Multimodal references
Up to 50 combined image, video, and audio references can provide appearance, location, movement, sound, and style cues. Clear roles help the model interpret them as one coordinated direction.

Feature 03 · Reference interpretation
Seedance 2.5 is designed to read a reference video's intention, framing, and cinematic language. A creator can direct pacing and camera behavior while changing the subject and story.

Feature 04 · Production control
White-model control, green-screen editing, camera direction, performance blocking, and a wider range of audio-video edits make the model more relevant to planned production workflows.

Prompt recipes
The fictional Last Tram project turns the four core capabilities into practical prompt structures. Save or adapt them now, then use them when Seedance 2.5 becomes available on Nanobana.

A clear opening, development, and resolution gives the model enough direction to pace one continuous sequence.
Create a continuous 30-second cinematic short at blue hour in a rain-washed coastal city. A cobalt-blue vintage tram arrives for its final route. Begin with a wide view through sea mist, track alongside the tram as it slows, move inside to an amber-coated conductor checking a silver case, then end as the tram disappears around the cliff road. Keep the tram design, wet stone streets, teal-and-amber grade, realistic motion, and quiet reflective mood consistent. Include natural tram sounds, rain, distant waves, and no on-screen text.
Explore what is coming
Give each reference a job so several creative inputs support one coherent scene instead of competing.
Use @Image1 for the exact cobalt tram design, @Image2 for the conductor's amber coat and face, @Image3 for the silver case, @Video1 for the restrained walking rhythm, and @Audio1 for the bell and wet-track ambience. Create three connected shots at the same coastal station. Preserve character identity, tram proportions, case details, weather, lighting direction, and color grade across every shot. Interpret the references as one cinematic world rather than arranging them as a collage.
Explore what is coming
A reference supplies pacing and framing while the subject, location, and story remain original.
Use @Video1 only as a reference for camera language: begin with a slow push through foreground reflections, transition into a lateral tracking move, then finish on a held close-up. Reinterpret that rhythm inside a vintage coastal tram at night. Follow the conductor from the rear door to a passenger holding a silver case. Do not copy the reference subjects or location. Keep movement deliberate, eyelines natural, reflections physically plausible, and the cobalt, teal, and amber palette consistent.
Explore what is coming
Define what must be replaced and what must remain untouched during a focused production edit.
Use @Video1 as the green-screen source. Keep the conductor, amber coat, silver case, body movement, timing, and camera angle unchanged. Replace only the green background with a rain-washed coastal tram platform at blue hour, including the same cobalt tram, wet stone reflections, sea mist, overhead practical lights, and distant waves. Match edge light, contact shadows, perspective, depth of field, and ambient sound so the composite feels like one photographed scene. Add no text or logos.
Explore what is comingHow it will work
The exact Nanobana controls will be finalized with the API integration. These steps focus on the creative decisions already supported by the model's published capabilities.
Describe the story beat, subject, environment, movement, camera language, sound, and how the 30-second segment should end.
Choose only the image, video, and audio materials that clarify identity, setting, motion, sound, or visual direction, and state each role.
After integration launches, review the first segment, extend only when the story needs it, then make focused edits without changing the whole direction.
Seedance 2.5 is a next-generation audio-video joint generation model developed by ByteDance. It is designed around 30-second storytelling, multimodal reference control, and more capable video editing.
Seedance 2.5 is developed by ByteDance. Nanobana is an independent creative platform preparing a web workflow for the model. Nanobana does not develop or own Seedance 2.5 and is not affiliated with or endorsed by ByteDance.
Not yet. Seedance 2.5 has been released, and Nanobana integration is coming soon. This page currently provides verified capability guidance and prompt planning without submitting generation tasks to another model.
The published model page describes a single generation of up to 30 seconds with the option to extend the video twice. Exact duration choices exposed by Nanobana will be confirmed when the API integration is complete.
Public model information describes up to 50 combined image, video, and audio inputs. Nanobana's final upload limits and accepted file requirements will be published with the integration.
A reference video can guide framing, camera rhythm, motion, and creative intent rather than acting only as a motion template. The prompt should still state which parts to interpret and which subjects or locations must be different.
ByteDance highlights white-model control, green-screen editing, professional camera movement, performance blocking, and broader audio-video editing requests. Nanobana will list the exact exposed controls after integration testing.
That is not confirmed yet. ByteDance's Seedance 2.5 page does not specify 4K output. Nanobana will publish verified resolution choices when its Seedance 2.5 integration is ready.

Use the prompt recipes to shape a 30-second sequence, choose purposeful references, and define the camera and edits before Nanobana generation opens.
Seedance 2.5 coming soon