A compact desk lamp sits beside an open sketchbook. The video begins close to its folded arm, then the camera moves back as warm light reaches the page. For a proposed startup launch film, I would build the opening around that single change. The viewer should understand what kind of object is being introduced and why it belongs on a working desk before any product description appears.
The hypothetical brand in this example would already have a finished lamp design and approved product photographs. My task would be to make a short launch video from those materials, not to invent a product or claim a successful campaign. I would want three generated scenes: a close-up of the lamp, a wider desk reveal and a calm final view with room for the actual product name.
The reference pack would include a front view, a side view and a photograph of the intended desk setting. I would choose one finish, such as a muted terracotta body, rather than mixing several colour variants in the same film. A brief self-recorded camera move could provide the motion reference, and an approved sound bed could establish a quiet, focused mood.
The first creative decision would be to keep the lamp still while the camera reveals it. Animating a hinge or switch introduces a more complicated mechanical action that may not be necessary for the launch concept. A small shift in viewpoint and a pool of light on the paper can show the form clearly. The film would describe an atmosphere of use without pretending to be a measured lighting test.
Bringing those references into Seedance 3.0 would make this an AI video-generation project. The model supports text-to-video, image-to-video and video generation guided by text, image, video and audio references together. For the startup’s proposed film, the product photographs establish the subject, the movement clip guides the camera and the written brief connects those inputs to the video I want to generate.
Make the opening recognisable before making it dramatic
My first image-to-video shot would show the lamp head and part of the arm, with the desk visible below. I would avoid an extreme macro view that could belong to almost any object. The lamp’s profile needs to be recognisable early, particularly when the finished clip is watched on a phone. Soft side light could reveal its finish while the camera moves just enough to show depth.
A possible prompt would ask for the supplied terracotta lamp on the same desk, a slow backward camera movement and warm light falling onto the sketchbook. It would specify that the existing arm, shade and base remain consistent. No extra switches, logos or accessories would be requested. This gives Seedance 3.0 a specific moving scene to generate rather than a vague instruction to make the product look impressive.
If the first result bends the lamp arm or changes the shade’s proportions, my revision would reduce the movement and use a more suitable side reference. A camera path that reveals too much unseen surface can make the product harder to preserve. I would prefer a shorter usable clip to a longer reveal with a distracting shape change. The launch film depends on the object staying recognisable from shot to shot.
Use a Seedance 3.0 motion reference to reveal the workspace
The second scene would widen the composition to include the sketchbook and a simple pencil holder. For a movement reference, I could record a smooth backward camera move across an ordinary table. The prompt would explain that only its speed and direction should be followed. The lamp photographs and desk image would define the objects and environment, so the reference clip does not become an accidental replacement setting.
This is a practical use of Seedance 3.0’s multimodal inputs. Text assigns a purpose to each reference, an image supplies the product’s appearance and a video communicates the desired camera behaviour. The generated output is a new video scene. It is not a product rendering delivered as a still image or an automated redesign of the workspace.
I would give the camera a clear endpoint: stop when the complete lamp and the lit part of the sketchbook fit comfortably in the frame. The pencil holder should remain secondary, with no animated hands or pages flipping themselves. The scene has enough movement in the camera and light. Adding unrelated action would make it harder to understand which object the startup is introducing.
Let the sound reinforce the video concept
The audio reference would be a short licensed or original instrumental passage with a gentle pulse. I would ask for an understated audiovisual scene that feels focused rather than theatrical. There is no need for a loud switch click if no switch operation is visible. Sound should support the lamp reveal, not imply a mechanical feature or product behaviour that the film does not establish.
Because Seedance 3.0 can generate audio with video, I would review the returned sound together with the picture. An unexpected voice or abrupt effect could change the tone of the whole launch film. A useful revision would ask for a quieter soundscape without speech, or the final edit could use the approved original track. The choice depends on what the actual generated clip contains and what the release requires.
I would also keep a version that makes sense with the sound muted. The visual progression should still be clear: a close view of the lamp, a reveal of the desk and a resting final frame. Short captions added during editing can identify the product without turning every moment into an advertisement. The moving images should carry the main introduction before text supplies the factual details.
Explore with text-to-video, then return to the real product
Text-to-video could be useful earlier in the creative process, before the desk setting is decided. I might describe a small creative workspace, warm task lighting and a quiet evening atmosphere to generate a moving mood study. That experiment could help choose between a bright studio setting and a more intimate desk scene. It would remain a concept, not footage of the actual lamp in use.
For the presentation version, I would bring the approved product images back into Seedance 3.0. The launch needs a particular lamp, not a plausible substitute generated from a broad description. I would compare its silhouette, finish and visible controls across the selected clips. If one shot differs noticeably, that shot should be revised instead of expecting a colour adjustment in the final edit to conceal the mismatch.
Give each additional Seedance 3.0 generation a purpose
A small launch-video budget benefits from specific attempts. I would first seek one acceptable opening and one acceptable desk reveal, then use the remaining generation allowance to resolve a visible problem. Before starting, the team should inspect the current model settings and the displayed generation cost. I would not estimate the film’s cost by assuming that every duration, resolution or reference combination consumes the same credits.
My review notes would describe changes the model can interpret: slow the camera, retain the shade profile, remove the extra object or stop before the desk edge fills the frame. “Make it better” gives the next attempt very little direction. A concrete revision also helps distinguish an issue in the source material from one created by an overly ambitious motion request.
End with a useful launch asset
The closing scene would hold the lamp in its final desk composition while the camera comes to rest. I would add the actual product name and an accurate invitation to learn more in a separate editor. Any specification, availability statement or price must come from the business’s real information. The generated video supplies the product presentation; it does not establish those commercial facts.
The finished deliverable would be a concise launch film and a silent captioned version, both built around the same three scenes. Seedance 3.0 would have generated the moving product imagery from text and audiovisual references, while the creator selects, revises and assembles the clips. For this startup concept, success means that viewers can follow a clear reveal of one lamp and finish with a recognisable impression of the product.