ByteDance · A long-take video model built for storytelling

Seedance 2.5

Cinematic single takes up to 30 seconds, with dialogue, sound effects and ambience generated in the same pass. Lock the composition with a first frame, pin your characters with @image1, preview a 480p draft and upgrade it to 1080p in one click — the generator is right below.

Seedance 2.5
Single takes, any length
4–30s
Maximum resolution
1080p
Including 21:9 and adaptive
7 ratios
Dialogue and SFX built in
Audio + video

What is Seedance 2.5?

Seedance 2.5 is the latest video generation model in ByteDance's Seedance family. It raises the length of a single generation to 30 seconds, so one shot can carry a whole beat of a story instead of being stitched together from short clips in an editor.

It understands text, images, video and audio as input. Write a paragraph and get a finished clip, upload a first and last frame to lock how the shot opens and lands, or give it several reference images and point to characters, outfits and locations with @image1, @image2 in your prompt. Dialogue, sound effects and ambience are generated natively, in sync with the picture.

On Moluri, Seedance 2.5 powers the AI video generator. You can also create a still you love with Nano Banana 2.1 and use it as the first frame for Seedance 2.5, bringing the composition to life.

Three ways to generate — which to choose

One model, three starting points. Picking the right one beats rewriting your prompt over and over.

ModeWhat you provideBest forAspect ratio
Text to videoA text descriptionExploring ideas from scratch7 options
First/last frameA first frame (plus optional last frame)Animating an existing image with a precise start and endFollows the image
Reference video1–9 reference images + a prompt using @image1, @image2…Keeping characters, outfits or products consistent7 options

Seedance 2.5 specs

Seedance 2.5 specs
SpecValueNotes
ModesText / first-last frame / referenceSwitch anytime under “Mode” in the generator.
Duration4–30 seconds, or autoLet the model choose the length — credits for 30 seconds are reserved and the difference is refunded.
Resolution480p / 720p / 1080pIterate at 480p, review at 720p, deliver at 1080p.
Aspect ratio16:9 · 4:3 · 1:1 · 3:4 · 9:16 · 21:9 · adaptiveFirst/last frame mode always follows the first frame’s ratio.
AudioDialogue + SFX + ambienceGenerated with the picture by default; turn it off for a silent video.
Reference imagesUp to 9 (in Moluri)JPG / PNG / WEBP / GIF, aspect ratio 0.4–2.5, 300–6000 px per side.
Draft mode480p draft → 1080p finalThe final keeps the draft’s prompt, references, duration and ratio. Drafts can be upgraded for 7 days.
SeedOptionalThe same settings and seed give similar results, which makes fine-tuning easier.
Web searchOptionalLets the model look things up — useful for real landmarks, current events and specific products.
Output formatMP4Download your video as soon as it’s ready.

How to use Seedance 2.5 in three steps

From one sentence to a finished clip with sound

Pick a mode

Text to video, first/last frame or reference video — choose based on what you already have.

Prompt shot by shot

Describe the subject, action and camera movement. Break long videos into timed beats and name references with @image1.

Generate and download

Most videos are ready in 1–5 minutes, with picture and sound together. Upgrade a draft you like to 1080p in one click.

The timed-beat prompt formula

Seedance 2.5 has enough time to tell a full beat of story. Write your prompt like a shot list and the model will play it in order.

Example prompt

[0–5s] An old canal town in morning mist. Static low-angle shot as a small wooden boat glides out from under a stone bridge. [5–12s] The camera tracks alongside the boat as the boatman rows, ripples spreading across the water, a rooster crowing in the distance. [12–20s] The camera slowly rises over the rooftops to reveal the whole town lit by the morning sun, cinematic color grade.

Subject + action
Say who is doing what before describing the setting — the model locks on faster.
Camera language
Low angle, tracking, push in, crane up — use film terms to control movement.
Timed beats
Markers like [0–5s] pace the action so events don’t pile up.
Describe the sound
Write the dialogue, effects or ambience you want to hear — audio is generated with the picture.

Seedance 2.5 FAQ

What is Seedance 2.5?

Seedance 2.5 is the latest AI video model in ByteDance's Seedance family. It generates single takes of up to 30 seconds and creates dialogue, sound effects and ambience natively alongside the picture.

How long can Seedance 2.5 videos be?

Anywhere from 4 to 30 seconds, set to the second. You can also choose “Let the model decide” and it will pick a length that suits the content.

Which resolutions and aspect ratios are supported?

480p, 720p and 1080p, in seven aspect ratios: 16:9, 4:3, 1:1, 3:4, 9:16, 21:9 and adaptive. In first/last frame mode the ratio follows your first frame.

Can I turn an image into a video?

Yes. First/last frame mode uses your image as the video’s opening frame (with an optional closing frame). Reference mode uses several images as references for characters, outfits or scenes — refer to them as @image1, @image2 in your prompt.

Do the videos have sound?

Yes, by default. Dialogue, sound effects and ambience are generated in sync with the picture. If you only need visuals, switch off “Generate audio” in the settings.

What is draft mode?

Draft mode renders a low-cost 480p preview first. If you like it, click “Generate the 1080p final” on the result card — the final keeps the draft’s prompt, references, duration and aspect ratio. Drafts can be upgraded for 7 days.

Am I charged if a generation fails?

No. Credits for failed or rejected jobs are refunded to your balance automatically.

How do I write the best prompts?

Write shots, not stories: describe the subject, action, camera movement and light, and break longer videos into timed beats. In reference mode, avoid words like “edit”, “replace” or “extend” so the job isn’t treated as a video edit.

Start creating with Seedance 2.5

Failed jobs are fully refunded, and everything you make is saved to your library.

See also:Nano Banana 2.1