AI game trailer generator: 4-30 s shots joined in Timeline
Build a game trailer from separate seedance-2.5 shots of 4 to 30 seconds, then join them in Timeline 1.0 with fade or wipe transitions between slots.

There is no single "trailer" endpoint. You generate each beat of the trailer as its own seedance-2.5 shot (4 to 30 seconds each), then cut the shots together with Timeline 1.0, which can put a fade, wipeleft or similar transition between slots.
ByteDance's launch post describes Seedance 2.5 as taking images, video clips and audio clips as references in one pass, which is how you keep a game's characters and look consistent across shots. Sume's side is from the video docs and Timeline docs, read 2026-10-01.
How long can each trailer shot be?
Sume's video docs say limits are not uniform across models: seedance-2.5 accepts 4 to 30 seconds. A trailer beat such as a logo reveal can be 4 seconds; a combat sequence can use up to 30. The docs name only a few models with ranges that long; check GET /v1/videos/models for the model you pick.
How do I keep the game's look the same across shots?
Send your key art, character sheets or a gameplay capture as input_references. The docs say audio and video references are honored by the Seedance 2.x models, so a gameplay clip can steer motion and a music file can steer rhythm. Only models whose supported_input_references lists a type accept it, so read the catalog entry first. For blocking out camera moves with a 3D model, see Seedance 2.5 clay render.
How do I join the shots with transitions?
Import each finished clip to media.sume.com with POST /v1/media-imports, then render a Timeline with one video[] slot per shot and an audio spine (audio.duration_seconds, 1 to 1800). Transitions go on slots after the first.
| Rule | Value in the docs |
|---|---|
| Where a transition may sit | On slots after the first |
| Types | fade, wipeleft, wiperight, slideup, slidedown, dissolve |
| Length | At most 1 s and at most 50% of the shorter neighbour |
| Too many fades | More than 8 adjacent fades is refused; insert a hard cut |
| Render needs | An Idempotency-Key header |
What should I do first?
Write the trailer as a shot list, generate the shots one request each, and run the unbilled Timeline /plan preflight before you pay for the render. Trailers usually want hard cuts on beats and a short transition only at tonal shifts; the 8-fade limit nudges you the same way.
Sources
Related posts
- AI movie trailer generator: shots, narration, music, titles
- Seedance 2.5 API: 30-second clips, resolutions, references
- How to assemble a long-form video with the Timeline 1.0 API
- Seedance 2.5 3D blockout camera control: clay renders as a reference
- AI game music generator: one track per level, cut to WAV
More in Use cases
- Multilingual video captions: language is a hint, not a font
On Sume video captions, `language` only tells speech-to-text what to expect. Look and font come from `style`, `design` and `font`, never from the language.
- ElevenLabs agent hold audio: 180 s, 40 MB; make it with Sume
ElevenLabs agent hold audio must be MP3 or WAV, up to 40 MB and 180 seconds. Steer a Sume track's length in the prompt, trim with Timeline audio, export mp3.
- FLUX Virtual Try-On v2: 4 MP inputs and a Sume reference edit
BFL's vto-v2 keeps inputs up to 4 MP as-is. Sume has no try-on endpoint in these docs; a garment swap is a reference edit with public HTTPS images.
- Seamless looping food video with AI: same first and last frame
To loop a food clip, send one image as image_url and end_image_url to Gemini Omni Flash 1.1, which Google says suits seamless loops. Clips run 3 to 10 seconds.
Written by Sume