Seedance 2.5 for Microdramas: 30-Second Video, Character References, and Dialogue Audio
Make a 30-second vertical microdrama with Seedance 2.5 on Pixocto. Set character and location references, generate dialogue audio, and adapt a complete scene prompt.
Seedance 2.5 for Microdramas: 30-Second Video, Character References, and Dialogue Audio
A 30-second drama needs connected shots, stable characters, and dialogue the viewer can hear. ByteDance's Seedance 2.5 overview highlights up to 30 seconds in a single audio-video generation and precise reference control. Pixocto's current workbench lets you choose 30 seconds, upload character and setting images, and turn on generated audio. This guide focuses on using those model and workbench capabilities; the script gives each one a job.
The Gallery works cited below are labeled Seedance without a version number. Their prompts illustrate ways to organize a drama, but they cannot be identified as Seedance 2.5 outputs from that label.
Generate one complete 30-second scene
Thirty seconds can hold one clear event without explaining an entire series. Put the timed shots in one prompt so entrances, props, and emotional changes share the same context. Pixocto's Seedance 2.5 workbench offers up to 30 seconds, leaving room for both dialogue and reactions.
The prompt attached to a nightclub romantic comedy divides its 30 seconds into six visible beats: seeing an ex, leaving, meeting a stranger, accepting an improvised rescue, growing closer, and ending on an open question. A late-night confrontation between spouses instead keeps a phone notification, a question, and a disturbing answer inside one living room. Both prompts specify actions for each beat instead of merely asking for rising tension.
Use reference images for characters and place
In the Seedance 2.5 workbench, choose reference images and upload separate images for the protagonist, the other character, and the setting. The current workbench allows up to 30 reference images, but two or three clear ones give a first two-person scene simpler roles: one for each person and one for the room. They are references, not three storyboard frames to play in order.
The prompt attached to a castle vampire romance assigns one image to the two leads' appearances and another to the castle environment. For your own prompt, state what each image controls, then name the facial, clothing, and spatial details that should stay stable. Reference images and first/last frames are different input modes in the workbench; this template uses references.
Generate dialogue, ambience, and silence together
The Seedance 2.5 workbench has a generated audio control. For a drama with speech, turn it on and specify who says each short line, plus the sound or silence after it. A long monologue leaves little time for reactions.
The prompt for the married couple's confrontation uses the wife's question, the husband's brief answer, and silence afterward. It also calls for a phone notification and quiet room tone. This is a useful way to write sound into the scene; check the actual output for lip movement, volume, and timing.
Set up the scene in Pixocto
| Control | Choice for this scene | Purpose |
|---|---|---|
| Model | Seedance 2.5 | Use the model covered here |
| Duration and ratio | 30 seconds, 9:16 | Leave room for speech and reactions in a vertical format |
| Input | Reference images for characters and setting | Give people and place clear visual anchors |
| Audio | Turn on generated audio | Include dialogue, a sound cue, and ambience |
Prepare the character and setting images, then replace every bracketed field in the prompt below. This is a new creative template, not the original prompt for any Gallery work or a verified output.
Create a 30-second vertical 9:16 live-action microdrama with generated audio. Use the uploaded reference images by role, not as sequential storyboard frames.
REFERENCE ROLES
The first uploaded image is the protagonist reference: [appearance and outfit]. Keep the same face, hairstyle, body proportions, and clothes in every shot.
The second uploaded image is the other character reference: [appearance and outfit]. Keep this person distinct from the protagonist throughout.
The third uploaded image is the setting reference: [room or place]. Keep its layout, lighting direction, and recognizable objects consistent.
SCENE
00:00–00:05 — In the reference setting, the protagonist notices [one concrete event or object that changes the situation]. Show it clearly before the reaction.
00:05–00:15 — The other character enters or responds. The protagonist takes [one visible action]. Maintain the same positions and screen direction across shots.
00:15–00:25 — The protagonist asks: "[one short question]." The other character reacts before answering. Keep the dialogue brief and natural.
00:25–00:30 — The other character says: "[one short answer that changes the meaning of the opening]." Hold the protagonist's reaction for a moment, then cut before any explanation.
AUDIO AND CAMERA
Clear, natural spoken dialogue, matching the speaker. Subtle room tone and [one motivated sound cue]. Leave a brief silence after the final line. Use restrained eye-level coverage, one two-shot, and close-ups only when a reaction matters.
AVOID
No extra characters, identity drift, wardrobe changes, moving props, reversed eyelines, invented on-screen text, long monologues, subtitles, or unexplained location changes.
After generating, check three things: whether faces and outfits stay consistent across shots; whether speech is clear and assigned to the right character; and whether the final pause gives the reaction enough time. A prompt can request these outcomes, but the footage still needs review.