Multimodal campaign shots
Combine approved appearance, movement, and sound references in one validated production request.
Video · Available
Seedance 2.5 multimodal video workflow
Seedance 2.5 AI video generator
Create Seedance 2.5 video with text, first-frame, and multimodal references using production duration, audio, format, and output controls.
- Provider
- ByteDance
- Base quote
- 112 credits
Multimodal input
01 / 03Give every Seedance 2.5 reference a job
A reference image can define appearance, video can define movement, and audio can define timing or sound. Upload only supported media and explain each role so the model does not have to reconcile an unstructured collection.
CreateForge verifies input ownership, type, size, and media duration where required. The client cannot substitute an unverified duration when it affects the request or cost.
Temporal planning
02 / 03Use the duration range for one coherent shot
Choose enough time for the action, camera path, and ending hold. A broader duration should not be filled with unrelated cuts or locations that reduce identity and continuity.
Configure resolution, audio, output format, references, and duration before quoting. The protected amount reflects the exact supported request.
Multimodal review
03 / 03Review visual, motion, sound, and format together
Compare reference identity, movement characteristics, and sound timing throughout the result. Inspect geometry after occlusion and camera turns, and confirm the clip reaches the requested endpoint.
Use the saved Library output and any returned final frame for the downstream workflow. Record which input or prompt layer needs correction instead of replacing the full multimodal brief.
Production-verified output
A real Seedance 2.5 provider acceptance result.
Public brief summary: one controlled continuous shot used to verify provider submission and video output handling; the complete acceptance prompt remains private.
- Verified
- 2026-08-17T07:18:20.610Z
- Public prompt hash
- 5f66491556b5…
- Acceptance settings
- text-to-video workflow · Adaptive · 480P · 4 seconds · MP4
Provider task succeeded with one video output during the documented production acceptance. This single production acceptance proves the connected task path, not universal motion quality, continuity, audio quality, or repeatability for every shot.
What Seedance 2.5 does inside CreateForge
Seedance 2.5 is a distinct production model with broader connected reference limits and its own duration, format, audio, and quote behavior. It is not a content alias for Seedance 2.0.
Use the mode that matches the source and keep each image, video, or audio reference purposeful. The production Blueprint validates media roles and duration before a quote is issued.
CreateForge tracks the task and saves eligible output in the private Library. Review multimodal fidelity, motion, audio, and output format before delivery.
Seedance 2.5 use cases
Choose the model for a specific deliverable.
Longer directed actions
Use the connected duration range for a continuous action that genuinely needs additional temporal room.
Format-aware delivery
Select the output container and optional final frame for a defined editorial or continuity need.
Generation controls
A consistent form, tuned to the model.
Prompt
requiredDescribe the video to create with Seedance 2.5.
Generation mode
requiredChoose the input workflow before adding reference media.
Text to Video · Frames to Video · Reference to Video
First frame
optionalAccepts up to 1 image file.
Last frame
optionalAccepts up to 1 image file.
Reference images
optionalAccepts up to 30 image files.
Reference videos
optionalAccepts up to 10 video files with a 30 second combined limit.
Reference audio
optionalAccepts up to 10 audio files with a 30 second combined limit.
Aspect ratio
requiredChoose a supported output frame with visual ratio cards.
Adaptive · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9
Resolution
requiredYour selected resolution is included in the credit price.
480P · 720P
Duration
requiredChoose 4–30 seconds in Video Settings.
Prompt blueprint
- 01Use multimodal references only when each file communicates a distinct visual, motion, or sound requirement.
- 02Describe one shot with a clear endpoint even when the connected duration range is broader.
- 03Choose MP4 or MOV for the downstream workflow and confirm it in the protected quote summary.
Multimodal product shot
Use the product images for exact shape and color, the reference video only for the slow clockwise movement, and the audio reference only for the soft mechanical rhythm. Keep the product centered as the camera pushes in, cool studio light, movement settles facing camera, no text or new objects.
Explicit roles prevent appearance, motion, and audio references from competing.
Longer environment reveal
Begin close on wet dark stone, then pull back steadily to reveal a compact pavilion beside the sea, one continuous camera move, architecture remains stable, low clouds travel slowly left, wind and distant surf audio, end on a wide composition and hold briefly.
Uses additional time for one camera reveal rather than several scenes.
Seedance 2.5 FAQ
Know the real product boundary.
Yes. The canonical model page exposes its production modes, multimodal input controls, and protected quote.
Use the supported reference roles shown by the production controls. Explain each source role and keep within the verified media limits.
The current connected workflow displays MP4 and MOV controls. Confirm the selected container in the quote summary.
They are separate adapters with different current input limits, duration and format controls, minimum credits, and quote behavior.