Turn a text prompt and optional image, video, or audio references into a generated video with sound, reference control, and focused editing options.
Create with Seedance 2.5
Choose the input you already have.
Keep the current output choices visible while you create.
Seedance 2.5 example
From Direction to a Finished Video
OVERVIEW
Use one creation tool to turn a prompt and optional references into a longer video with coordinated picture, movement, and sound.
Seedance 2.5 is an AI video generator for stories that need more than a brief visual beat. Enter a written scene, then add image, video, or audio references when a subject, movement, rhythm, or sound needs firmer control. The tool combines those inputs into a generated video with audio, so the page leads directly from an idea to a creation task.
The workspace is intended for filmmakers, creative teams, and video creators who want one place to enter the core request and the references that shape it.
The model can create up to 30 seconds in one generation and can extend a result up to two times. This duration supports tasks that need several related visual beats instead of one very short clip. It also gives creators a single generated sequence to refine when the first result needs a focused change.
Reference control lets each upload do a practical job. A source video can guide framing, camera movement, pacing, or blocking. Images can establish the look of a subject or scene, while audio can guide rhythm, performance, or sound. After generation, a focused edit instruction can change a selected visual or sound element without replacing the rest of the request. You stay in a direct input, generate, and refine workflow.
Seedance 2.5· 30-second stories· reference control· video and sound
CONTROL
Move from a written scene to a longer, reference-led video while keeping each creative instruction tied to a visible result.
You can generate a video up to 30 seconds long in one pass, giving the action enough time to establish a place, develop a beat, and reach a clear ending. When the story needs more room, extend the result up to two times while continuing the existing direction.
Combine a text prompt with image, video, and audio references. Each input can carry a different part of the direction: appearance from an image, movement or framing from video, and rhythm or sound character from audio. The tool brings those instructions into one generated sequence.
Use a reference video to control more than copied motion. Seedance 2.5 can follow framing, shot rhythm, camera movement, and the broader cinematic intention. This helps you carry a deliberate visual language into a new subject, environment, or style.
Change selected elements in an existing video direction without rewriting the entire scene. You can ask for a different setting, prop, wardrobe detail, or visual treatment while preserving the rest of the requested composition, making focused changes easier to communicate.
You can direct audio and visuals as parts of the same sequence. The tool supports generated sound as well as requests such as removing music or adding narration that fits the picture. This gives spoken lines, ambience, rhythm, and action a shared creative brief.
You can use white-model references to guide camera movement, pacing, framing, and subject paths, or keep foreground action from green-screen footage while replacing the environment. These controls turn previsualized movement and isolated performances into finished scene directions.
RANGE
See how the tool moves from your prompt and references to a generated video, then accepts a focused follow-up edit.
Enter a prompt, then attach only the image, video, or audio references needed for the task. An image can define appearance, a source clip can guide movement, and audio can establish timing or sound character.
Seedance 2.5 brings the prompt and selected references into one video task. The requested picture and sound share the same input context instead of being described as separate, unrelated outputs.
Use the generated video as the next input when one detail needs work. Request a background, prop, wardrobe, narration, or music change while keeping the rest of the sequence focused on the original task.
Read the complete Seedance 2.5 guide for detailed prompt structure, reference roles, timing, sound direction, and editing examples.
FOCUS
Choose one tool when the result depends on longer duration, clear reference roles, coordinated sound, and targeted visual changes.
Generate up to 30 seconds in one task when several related visual beats need more time than a brief clip. The model also supports up to two extensions when the requested sequence needs additional duration.
Use images for appearance, video for movement or framing, and audio for timing or sound. Clear reference roles make the request easier to control and reduce conflicts between inputs.
Ask for a focused visual or sound edit without rebuilding the complete request. Change a background, prop, wardrobe detail, narration, or music choice while retaining the elements you want to keep.
Use white-model motion and camera timing, or retain a green-screen performance while replacing its environment. The tool can carry those source actions into the generated result instead of treating the uploads as decorative references.
SCENARIOS
Match each creative task with the prompt, reference material, and final details that matter most to that situation.
Use a prompt to describe the brand subject and the sequence you want to generate, then add product or environment images for appearance. This task benefits from the available duration when several related visual beats belong in the same request.
Describe the characters, location, visible action, camera distance, and intended sound. Add character images when appearance matters. The tool can use those references and the prompt together without requiring a separate visual request for each character.
Upload source footage with the product or foreground action, then request a different environment around it. Use the editing instruction to identify the background change and the source elements that should remain in the task.
Provide audio as a sound reference and images for the requested appearance or styling. State what each source contributes, then describe the visible action and camera direction in the prompt without asserting an unverified synchronization result.
Provide a white-model video as a reference for movement, framing, pacing, and subject paths. Add separate image references for the requested vehicles, characters, or materials, so production structure and appearance have distinct input roles.
Begin with green-screen footage that contains the foreground performance. Request a replacement setting while identifying the subject and action to retain. The background description can cover environment, depth, atmosphere, light direction, and motion.
METHOD
See input, editing, audio, and duration details for Seedance 2.5 alongside three current AI video models.
| Capability | Seedance 2.5 | Runway Gen-4.5 | Veo 3.1 | Luma Ray3.14 |
|---|---|---|---|---|
| Text-to-video generation | Supported | Supported | Supported | Supported |
| Image-guided generation | Image references | Image to Video | Reference images | Partial — Image to Video, no reference feature |
| Video-reference editing | Video references and focused edits | Not documented | Input-video controls | Modify Video |
| Native generated audio | Video with sound | Not documented | Native audio | Not supported |
| Longest documented duration | Up to 30 seconds in one generation | 2–10 seconds | 8-second clips with extension | 5–10 seconds; extension up to 30 seconds |
| Audio reference as input | Supported | Not documented | Not documented | Not documented |
ANSWERS
Get direct answers about inputs, duration, reference control, editing, sound, and the kinds of scenes the workflow can support.
Seedance 2.5 is an audio-video generation model for creating a video from a prompt and optional visual or audio references. The tool supports longer storytelling, reference-led control, and focused visual or sound edits. A single brief can therefore describe both what viewers see and what they hear.
A single generation can produce up to 30 seconds of video. The model also supports extending a result up to two times when a story needs additional development beyond the first generated sequence. Plan connected beats within that documented duration instead of assuming an unlimited continuous scene.
You can direct the task with text plus image, video, and audio references. Give each reference a clear purpose, such as subject appearance, camera language, motion, timing, rhythm, or sound character. Clear roles help keep several sources aligned around one requested result.
Yes. It is designed for joint audio-video generation, so sound can be part of the same creative direction as the picture. Your brief can include ambience, rhythm, narration, or other relevant sound cues. Describe those cues beside the visible action they are meant to accompany.
You can request focused visual changes such as a different background, prop, wardrobe detail, or treatment while keeping other elements unchanged. Clear, narrow instructions help communicate which parts should move and which should stay consistent. Name the retained elements as clearly as the requested change.
Reference control can carry intention, framing, camera movement, cinematic language, subject appearance, motion, or sound into a new generation. The goal is creative direction based on the role of each source, not only direct motion copying. Different source types can contribute separate parts of the same sequence.
A white-model or previsualization video can guide camera movement, shot rhythm, framing, subject motion, and blocking. Visual references can then add the detailed appearance of characters, objects, materials, or environments. This separates production movement from the final visual identity assigned through other references.
Yes. Green-screen editing can keep the foreground subject or performance and replace the surrounding environment from a written description. Your direction should cover depth, lighting, atmosphere, and details that connect the new world to the retained action. Identify the foreground action that should remain before describing the replacement setting around it.
Open the Seedance 2.5 guide for a detailed sequence covering prompts, reference roles, timing, controlled edits, sound direction, and output checks. Use the shorter steps on this page when you only need a quick start.
Enter your scene, add the references that matter, and start creating. For more help, open the full guide, or visit the pricing page.
