Seedance 2.0 prompt guide: how to write shots that work

Seedance 2.0 reads a prompt best when it describes one shot the way a director would brief a camera crew: what is in front of the lens, what happens, where it happens, how the camera behaves and what the image should look like. Writing in that order keeps the important parts at the start, where they carry the most weight.

Here is how a weak prompt turns into a strong one. "A coffee commercial" gives the model almost nothing to work with. "A barista pours latte art" adds a subject and an action. "A barista in a green apron pours a leaf of latte art into a white cup on a marble counter" adds detail the model can draw. Add the camera and the look and the shot is fully directed.

The shot size tells the model how much of the subject to show. The angle changes how the subject feels. Name both when framing matters.

Two prompts with the same subject and camera can feel completely different because of the light. Say where the light comes from, what colour it is and how hard it is.

Put the style at the end of the prompt. A short, concrete phrase does more than a list of adjectives: "35mm film, shallow depth of field" is clearer to the model than "beautiful, stunning, amazing, high quality".

Clips run 4 to 15 seconds. That is enough for one clear beat, not a whole story. A good test: could the action happen in one breath? "She opens the box and smiles" fits. "She wakes up, gets dressed, walks to work and opens a box" does not, and the model will either rush it or skip parts.

If the action is slow, ask for slow motion or a longer clip. If it is quick, a short clip with a clear camera move keeps it from feeling empty. For a story, plan a short list of shots and make one clip per shot.

Words are good at describing a type of thing. They are bad at describing one exact thing, such as your face, your product or your logo. For those, attach references: up to 9 images for appearance, 3 video clips for movement and 3 audio clips for timing or voice. Then refer to them in the prompt: "the woman from the first image", "the camera move from the clip".

Keep the prompt focused on what the references do not show: the action, the setting, the light and the camera. The Seedance references guide covers each type in detail.

Seedance 2.0 can generate sound with the clip at no extra credits. Describe it in the prompt like any other part of the scene: the ambience (rain on a window, a busy street), sound effects (footsteps on gravel, a door slamming) and the mood of any music.

For a spoken line, put the words in quotes and say who says them. Keep it short: one or two sentences fit comfortably in a single clip. For longer narration, generate the video and add a voiceover afterwards.

Each prompt makes one continuous shot, so a longer video is a sequence of clips. Write a short shot list first (wide establishing shot, medium shot of the action, close-up of the detail, final reveal) and give each clip its own prompt.

Keep the parts that must match identical in every prompt: the same description of the character, the same location and the same light. Copy those sentences word for word and change only the action and the camera. Add the same character reference to every clip, then join the clips in the free video editor.

Change one thing at a time so you know what fixed it.

Stacking several scenes in one prompt. Seedance 2.0 makes one continuous shot; if you need a cut, make two clips or use a multi-angle template.

Describing feelings instead of visuals. "An emotional moment" gives the model nothing to draw. "Her eyes fill with tears as she reads the letter" does.

Leaving the camera out. Without a camera instruction the model picks one for you. If the move matters, name it.

Asking for two camera moves at once. "Orbit while zooming in and tilting up" usually gives a muddle. Pick the one that matters.

Writing what you don't want. Describe what should be in the shot instead: "an empty street" works better than a list of things to leave out.

The free Seedance prompt builder puts these steps into a form: type the subject, the action and the setting, pick a camera move, a shot, the light and a style, and it writes the prompt in this order. It runs in your browser and shows the credit cost of the clip next to the prompt.

Aspect ratio is a setting, not a prompt word: pick 16:9, 4:3, 1:1, 3:4, 9:16 and 21:9 in the studio. Choose 9:16 for Reels and TikTok, 16:9 for YouTube and 21:9 for a cinematic letterbox.

Resolution changes the cost. A 5-second clip costs 15 credits at 720p, 30 at 1080p and 70 at 4K. Draft at 720p or with Seedance 2.0 Fast (12 credits for 5 s at 720p), then render the keeper at a higher resolution.