SEEDANCE 2.5
AI VIDEO GENERATOR
One workspace for text to video, multimodal reference, video editing, video extension, and first / last frame generation. Up to 30 seconds per clip, with the exact credit cost confirmed before every run.
Step 1 — Prompt
Describe the video you want before generating.
Your generated video appears here.
Set a prompt and output options, then hit generate.
KEY FEATURES AND LIMITS
This video generator extends single-pass generation from 15 to 30 seconds, and every mode below runs through one workspace with the cost shown before you submit.
30-second single pass
Generate up to 30 seconds in one submission, up from the 15-second ceiling on Seedance 2.0. Scenes, shot changes, and pacing develop inside that window instead of being stitched from separate clips.
Text, image, and first-last frame
Describe a scene in natural language, animate a still, or pin one opening frame and one closing frame and let the model interpolate the motion between them.
Up to 50 multimodal references
Combine 30 reference images, 10 reference videos, and 10 reference audio clips in a single job. Tag an image as the first frame or last frame to lock the composition.
Audio-only reference
A reference job does not need an image or a video. Upload audio alone and the model shapes pacing and motion around the sound.
Edit and extend existing footage
Attach a clip and describe the change in natural language, or ask to continue past the last frame or before the first, keeping subject and style consistent.
Cost and format decided up front
Pick 480p, 720p, or 1080p and an MP4 or MOV container, and see the exact credit cost recalculated before every run. 4K is not offered, so this workspace does not list it.
HOW TO USE STEP BY STEP
Three steps from an idea to a finished clip.
Describe the subject, the movement, the camera, and the mood. In a reference job, address your uploads with @image1, @video1, or @audio1.
Choose a duration from 4 to 30 seconds or let the model decide, pick 480p, 720p, or 1080p, then upload any combination of up to 30 images, 10 videos, and 10 audio clips.
Submit the task and watch the result panel. Long clips are asynchronous, so the panel keeps polling until the video is ready, then plays and downloads it.
WHO THIS WORKSPACE IS BUILT FOR
This workspace fits work where a longer, reference-driven take matters more than a single short clip. These are the teams that get the most out of it here.
Short-form creators
Ship a full 9:16 or 16:9 take in one pass instead of stitching three clips, and keep the character stable across every episode of a series.
Ecommerce and DTC brands
Keep product shape, colour, and finish accurate across a whole catalogue. Reuse the same approved references for listings, ads, and every market.
Marketing and ad creative
Hold brand look steady while changing only the part that needs to change, so a campaign can run many variants from one approved set of references.
Short drama and narrative teams
Cover a complete story beat inside 30 seconds. Character and costume references survive the cut, and extend keeps the same cast for the next scene.
Concept and previs teams
Turn static references into a moving draft to review with stakeholders before committing to a shoot, a render, or an expensive pickup.
Education and training
Explain a procedure, a safety step, or an abstract idea with a scene instead of a slide, and update the footage without re-shooting the lesson.
WHAT USERS SAY ABOUT THIS TOOL
Feedback from creators and teams running longer, reference-driven work.
使用midjourney和Seedance2.5制作 双刀一出，眼睛先认输。想象中的武侠决战 提示词在评论区
澜森
@lansenai
カメラに向かって、ただ遊ぶだけ その１６ Seedance2.5
keshi
@keshiAIart
991カット作って、画面に出たのは152カットでした。テイクは2,288本作って、採用は160本。この採用率を向上させるためにしてみたこと② 動きを先に作ってみる→生成。ベースで使用した動画モデルはSeedance2.5
志村優 | GOKKO
@ucmulleryu
This AI video is absolutely amazing 🔥 Seedance 2.5
Lex
@lexx_aura
劲爆金曲MV：《Safety & First Place》 歌词：GPT-6-Astra；音乐模型：Suno-V6；视频模型：Seedance2.5
Xanderwow
@Xanderwow_A
Introducing Seedance 2.5, our new-generation video creation model! With advancements in long-form storytelling, multimodal reference, and editing, Seedance 2.5 goes beyond longer single-pass video generation, delivering the journey from idea to finished video with greater control.
ByteDance Seed
@ByteDanceSeed_
SEEDANCE 2.5 VS SEEDANCE 2.0
Only the differences confirmed by the provider contract are listed here.
Maximum duration per clip
Auto duration
Reference images
Reference videos
Reference audio
Audio-only reference
Resolutions
Output container
Watermark option
SEEDANCE 2.5 FAQ
Common questions about the request contract and this workspace.
A single request renders up to 30 seconds. Durations from 4 to 30 seconds are accepted, and the automatic mode lets the model pick the length while billing the 30-second ceiling.
480p, 720p, and 1080p are supported. Seedance 2.0 offered 4K, but Seedance 2.5 does not and rejects it. Output is MP4 by default, with MOV available for editing and extension work.
Up to 30 images, 10 videos, and 10 audio clips. Reference videos may total 30 seconds across all clips, and audio references may also total 30 seconds. A job with only audio references is valid.
Yes. Assign the first-frame and last-frame roles to uploaded images. These roles require the adaptive aspect ratio and cannot be combined with video or audio references, because the model turns them into plain references whenever other media is present.
Editing takes a source clip of 4 to 30 seconds and follows its length, so the duration must be automatic and the ratio adaptive. Extension asks for more footage before or after the source clip. Both require at least one reference video.
The price depends on resolution and duration and is configured by the operator. The workspace shows the exact cost before you submit, and the server recalculates it and charges that amount. Jobs that fail are refunded in full.
ARTICLES ABOUT SEEDANCE 2.5
Practical guides for planning prompts, reference media, and output settings.
CREATE WITH SEEDANCE 2.5
Text, images, video, and audio references in one workspace, with the credit cost confirmed before every run.