Stories & performance
Official creative film
A creative film illustrating how separate shots and sound contribute to a complete sequence.
Official example · ByteDance SeedSEEDANCE 2.5 · AI VIDEO CREATION
Seedance 2.5 Turn a visual idea into a moving scene
Direct action and camera movement with text. Guide the result with image, video and audio references, or define its first and last frames. Generate video with native sound in one workspace.
- Up to 30 seconds
- Up to 1080p
- Native audio · optional
Review results, manage assets and access generation history in AI Studio.
See Seedance 2.5 in motion
Explore camera direction, connected action and sound design in published work, then choose an approach for your project.
Original sources are credited. Viewing notes are written by ImgAI.
Stories & performance
Backstage to stage
Follow a singer from backstage to performance; observe camera movement connecting different spaces.
Official example · ByteDance SeedEditing & continuation
Subway chase extension
A continued subway chase demonstrates how an extension picks up existing characters and action.
Official example · ByteDance SeedMultimodal references
Peking opera references
Character and scene references guide opera staging, with movement and occlusion shaping camera transitions.
Official example · ByteDance SeedMultimodal references
Concert ensemble
Separate references define venue and performers; examine how an ensemble is arranged across shots.
Official example · ByteDance SeedMultimodal references
Fantasy from a clay render
A clay render supplies blocking while image references guide the fantasy setting and materials.
Official example · ByteDance SeedEditing & continuation
Green-screen football
Green-screen movement is placed in training and match settings; observe continuity around the performer.
Official example · ByteDance SeedEditing & continuation
Breakfast camera editing
Camera edits reshape a breakfast sequence while retaining its subjects and actions.
Official example · ByteDance SeedEducation & industry
Poetry classroom
A Doubao Classroom example turns poetry into moving imagery linked to the lesson.
Official example · ByteDance SeedEducation & industry
Car assembly demonstration
Clay-render references define component positions and assembly order, with added materials and lighting.
Official example · ByteDance SeedEditing and extension examples illustrate model capabilities. See the specifications below for creation modes currently available on ImgAI.
How creators combine their tools
These X posts combine GPT Image 2.5, Seedance 2.5 and other tools. Open the original posts to explore each creator’s process.
From product stories to short films
Plan references and shots around the deliverable to give the model a clear, actionable direction.
Product advertising
Use a product image to define appearance, then describe a complete demonstration such as rotating, opening or spraying. Specify which markings, shapes and materials should remain consistent.
Fashion and character films
Describe styling, setting and movement separately. Use reference images for appearance, then direct blocking, expression changes and camera distance in the prompt.
Narrative and storyboard previews
Set out events and shot transitions in sequence. Validate the key action before building a longer passage, keeping each generation focused on a manageable story beat.
Music and dance visuals
Supply audio or movement references and specify whether they should guide rhythm, action or camera work. Give the performance, music and ambience distinct roles.
Architecture and spaces
Use images to establish the space, then direct an entrance, walkthrough or orbit. Define camera direction and speed instead of relying on vague cinematic adjectives.
First-to-last-frame transitions
With an opening and closing image ready, use first/last-frame mode to direct what happens between them. Explain how the subject reaches the final position.
Seedance 2.5 specifications and credits
Compare the resolutions and durations available on ImgAI before choosing your generation settings.
- Creation modes
- Text to video, image first frame, multimodal references, first and last frames
- Reference images
- Up to 30 in multimodal reference mode
- Video and audio references
- Up to 10 videos totaling 30s; up to 10 audio clips totaling 30s.
- Aspect ratios
- 16:9
- 9:16
- 1:1
- 4:3
- 3:4
- 21:9
Adaptive aspect ratio is available and is used for first-frame and first/last-frame generation.
| Resolution | 5s | 10s | 15s | 30s |
|---|---|---|---|---|
| 480p | 687 | 1,373 | 2,059 | 4,118 |
| 720p | 1,543 | 3,086 | 4,629 | 9,258 |
| 1080p | 2,750 | 5,499 | 8,248 | 16,495 |
With video references, credits depend on the combined reference and output duration at the selected resolution. Image and audio references do not add video seconds; the audio toggle does not change these prices. Check the composer total before submitting.
View credit plansWrite a prompt the camera can follow
Specify the subject, action sequence, camera and sound. The snippets below illustrate prompt structure; they are not the original prompts for the videos above.
Define the main action
Establish who does what, and where. Add materials, lighting and style after that. A single clear action makes the result easier to evaluate.
Example wording
A clear perfume bottle stands on a black surface. The camera slowly pushes in as a rear side light passes through the glass. Keep the background minimal.
Sequence the action
For longer clips, use a few time ranges to establish order. Give each range a visible change and allow time for movement to unfold.
Example wording
0–3s: the character walks to the window. 3–7s: they stop and open the curtains. 7–10s: push toward their profile as daylight enters the frame.
Separate subject and camera movement
Specify the subject’s action independently from the camera path. Include framing, direction and speed, and avoid contradictory camera instructions.
Example wording
The subject stays in place and slowly turns their head. The camera moves a quarter-circle from the front toward their right side, keeping a medium close-up and constant distance.
Assign a role to each reference
Use the composer’s reference mentions to say which assets define appearance, setting, movement or sound. Avoid giving a single reference conflicting roles.
Example wording
Use the first image for the character and the second for the setting. Borrow only the camera timing from the video, preserving the clothing and background shown in the images.
Direct the sound separately
When audio is enabled, describe dialogue, ambience and music separately. Identify speakers and their exact lines; explicitly state when music or speech is unwanted.
Example wording
A soft glass tap as the bottle touches the surface. Quiet ambient music in the background. No dialogue and no voice-over.
Test short clips, then refine
Start by testing framing, the main action and reference roles. Retain effective instructions and change one key variable at a time before increasing duration or resolution.
Example wording
Keep the character, lighting and action sequence. Change only the slow camera push to a locked-off shot; retain all other instructions.
Seedance 2.5 FAQ
Start with text, upload an image as a first frame, use multimodal references, or provide first and last frames. Choose the mode that matches your assets, then set duration, resolution and sound.
ImgAI currently offers 5, 10, 15 and 30 seconds at 480p, 720p or 1080p. Reference assets and creation mode can also affect the available aspect ratios; follow the options shown in the composer.
Reference images guide the appearance of characters, objects or settings without defining the opening frame. First and last frames specify how the video begins and ends. Use the matching creation mode for each.
Yes. Enable audio and describe speech, ambience or music in your prompt, or disable it for silent video. The audio toggle does not change the current credit price.
With video references, the total reference and output seconds are charged at the reference-video rate for the selected resolution. The table for generations without video references does not apply; the composer shows the task total before submission.
This page features already-published videos with source credits beside each example. Viewing notes and prompt guidance are edited by ImgAI; they are not claims about the creators’ original prompts or guarantees of identical output.
Start with one well-defined shot
Prepare your references and test action and framing with a short clip before adjusting resolution and duration.