Short answer: Use Media to Video when a single prompt needs more than one kind of reference. Give each uploaded asset one job—appearance, motion, pacing, or environment—then use the labels shown in the interface to assign those jobs. More references do not automatically make the result more controlled.
When Media to Video is the right workflow
Choose Media to Video when text alone cannot define the visual result. A practical setup might use:
- an image to define a product, person, or starting design;
- a video to suggest motion, camera direction, or pacing;
- an audio file to guide pacing, mood, or a visible beat;
- a written prompt to define the subject, setting, action, and ending.
If you only need to animate one still image, Image to Video may be simpler. If you are describing a scene from scratch, Text to Video may be clearer. Media to Video is most useful when the references have complementary roles rather than overlapping instructions.
Open the workflow and check the inputs
Open Media to Video, select Seedance 2.5, and choose the reference mode available in the interface. Upload the image, video, or audio assets you actually need. Use the asset labels shown in the interface—such as @Image1, @Video1, or @Audio1—instead of assuming labels from another generation will carry over.
Before generating, review the output settings and estimated credits. The current creator exposes duration presets from 4 to 30 seconds, six aspect ratios, 480p, 720p, and 1080p output, and optional audio controls. These settings affect how much motion and timing can reasonably fit into the brief.
Assign one role to each reference
Start with a small input map:
| Asset | Role to state in the prompt | Avoid asking it to control |
|---|---|---|
| Image | subject appearance, color, shape, or starting frame | every camera move and edit |
| Video | motion language, camera direction, or pacing | the identity of a different subject |
| Audio | rhythm, atmosphere, or a timing cue | exact preservation of the source track |
If two assets describe the same property differently, choose one authority. For example, let the image control the mug’s appearance and let the video control the camera move. Do not ask both to define the product design.
A self-contained prompt pattern
This is an illustrative prompt, not a tested result. Prepare a reference image, a short motion excerpt, and an audio excerpt with one clear accent and a sustained ending:
Use @Image1 for the travel mug's appearance, @Video1 for one slow camera move,
and @Audio1 only for pacing and mood. Do not copy unrelated subjects or objects
from any reference.
Create a 10-second 16:9 product video on a light wooden table beside a window.
Keep the mug's shape, lid, handle, color, and visible markings consistent with
@Image1. 0–3s: stable medium shot with the mug centered. 3–7s: make one slow
left-to-right move inspired by @Video1 while the mug remains still. 7–10s: ease
to a stop and hold the final frame while the audio resolves.
Preserve markings already visible in @Image1, but add no new text or logos. Use
@Audio1 to guide pacing and mood; exact reproduction of the source audio is not
guaranteed. No hands, extra products, zoom, orbit, or cuts.
The prompt makes the hierarchy explicit: @Image1 is the appearance reference, @Video1 is the motion reference, and @Audio1 is a timing and atmosphere reference. If the interface uses different labels, replace them with the labels displayed after upload.
Prevent reference conflicts
Reference conflicts usually come from assigning the same decision to several inputs. Use these checks:
- Appearance conflict: If the image shows a red mug and the motion video shows a blue bottle, state that the image controls the mug and the video controls motion only.
- Camera conflict: If the text says “locked camera” but the video reference contains a tracking move, remove one instruction.
- Timing conflict: If the audio has several peaks, identify the one visible action that should land on the main accent instead of asking the clip to follow every beat.
- Ending conflict: State whether the last frame should hold, fade, or finish on a particular composition.
Shorter prompts are easier to debug. Remove decorative adjectives before removing the subject, timing, camera, or ending.
Review the result in the right order
First check whether the subject and framing match the image reference. Then check whether the camera movement resembles the intended motion reference. Finally compare the visible action with the timing cue from the audio.
An audio reference does not mean the original audio track will be preserved exactly. If the final video must keep the original music or narration, import the generated video and source audio into an external editor with a waveform and align them there.
If one input caused a problem, run a focused retest: keep the image and prompt fixed, then remove or replace only the motion or audio reference. Changing every asset at once makes the cause harder to identify.
FAQ
Can I upload many references for more control?
The current product page describes support for multimodal references, including images, video, and audio, with up to 50 reference assets. Use only the assets that add a distinct constraint; a larger set can introduce competing directions.
Does the video reference copy its subject into my result?
Not by default if you define its role clearly, but reference behavior can vary. State that the video controls motion or camera language only, and explicitly describe the subject that must come from the image reference.
Can audio alone define the visual story?
An audio reference alone does not reliably specify the visual story you want. Add a visual prompt to define the subject, setting, action, camera, and ending.
Which workflow should I use for one image?
Use Image to Video when one image is the main control and no additional motion or audio reference is needed. Use Media to Video when multiple reference types have separate jobs.
For model context, see the Seedance 2.5 model overview. For planning timed shots, read Seedance 2.5 Storyboard Prompts.
