Seedance 2.5 AI Video Generator

Describe the shot, or drop in reference images, video and audio, and Seedance 2.5 renders picture and sound together — up to 30 seconds in a single generation, right in your browser.

Drag and drop or upload images, videos, or audio as reference (optional)Guidelines

0 credits

Made with the Seedance AI Video Generator

Four clips straight out of the generator - no edits, no colour grading, no post-production.

Multi-shot consistency

A chrome-limbed performer, a circus ring, a butterfly throne - three shots in ten seconds, and the character survives every cut.

Cinematic camera work

The camera crawls in from a wide corridor to a close reaction shot, holding low light and real weight all the way through.

Dialogue with lip sync

She speaks straight to camera in a torchlit barn, and the mouth movement lands on the generated audio, not near it.

Any art direction

Same generator, hand-drawn anime look: a cloak whipping in the rain above a neon-lit city, linework and all.

Key Features of the Seedance 2.5 AI Video Generator

Bytedance’s flagship video model, built to generate picture and sound in a single pass instead of stitching them together afterwards.

Image, Video and Audio References

Anchor a shot with several kinds of material at once. Images carry identity, style and composition; a reference video carries camera motion and action; reference audio carries vocal timbre and melody. Point at each one by index in the prompt — “use the composition of Image 1”, “match the motion of Video 2”.

Every Seedance Model in One Place

Switch between Seedance 2.5 for the shot you publish and Seedance 2.0, 2.0 Fast or 2.0 Mini for cheap drafts. A 6-second 720p draft on the fast models lands in roughly 25 seconds, so you can block out the idea before paying for the full-quality render.

30 Seconds in One Generation

Single-generation length doubles from 15 to 30 seconds, with smoother transitions between shots and a character who still looks like themselves at the end of the clip. Pick any length from 4 to 30 seconds per render.

1080p, Vertical and Ultrawide

Render at 480p, 720p or 1080p and frame it for where it is going: 16:9 for YouTube, native 9:16 for Reels, Shorts and TikTok, 21:9 for a wide cinematic cut, plus 1:1, 4:3 and 3:4.

Dialogue, Sound and On-Screen Text

Sound is generated in the same pass as the picture, so a spoken line lands on the mouth instead of near it. Write the line in the prompt and ask for it back as voiceover, a lip-synced reply or a speech bubble — and ask for subtitles if you want it on screen too, timed to the audio.

Edit, Extend and Stitch Clips

Upload a video and change it by prompt: replace the main subject, add or remove an object, or repaint a specific moment by timestamp. You can also extend a clip forwards or backwards, or join up to three clips into one continuous scene.

Frequently Asked Questions

What is the Seedance 2.5 AI video generator?

Seedance 2.5 is Bytedance’s flagship video generation model, and this page is a browser front end for it. You describe a shot in plain language, optionally attaching reference images, video or audio, and the model renders the picture and the soundtrack in a single pass. It reads scripts with cinematic precision, keeps characters and scenes consistent across cuts, and can edit existing footage as well as generate new shots.

How long can a Seedance 2.5 video be?

Anything from 4 to 30 seconds in one generation — double the 15-second ceiling of Seedance 2.0. For something longer, upload the finished clip and ask the model to extend it forwards or backwards, or to stitch up to three clips into one continuous scene.

How fast is generation?

It depends on the model, the length and the resolution. A 6-second 720p draft on Seedance 2.0 Fast or Mini is usually ready in about 25 seconds; a 30-second 1080p Seedance 2.5 render takes considerably longer. Generation continues while you keep working, and your recent tasks come back when you reload the page.

Which aspect ratios does it support?

Choose 16:9, 9:16, 4:3, 3:4, 21:9 or 1:1 before you generate, or leave it on auto and let the reference material decide. 9:16 is native rather than cropped, so vertical cuts for TikTok, YouTube Shorts and Instagram Reels fill the frame, and 21:9 gives you a wide cinematic ratio for a trailer.

What resolution and file format do I get?

Seedance 2.5 renders at 480p, 720p or 1080p and downloads as an MP4 with the audio already mixed in. Iterate at 480p while you are still working on the prompt — it is the cheapest option — then re-run the version you like at 1080p for publishing.

Does Seedance 2.5 generate sound?

Yes, and it is generated with the picture rather than added afterwards. That covers spoken dialogue with matching lip movement, voiceover narration, ambient atmosphere and sound effects tied to the action on screen. Write the line you want in the prompt; add a subtitle instruction and it appears on screen, timed to the audio.

What can I upload as a reference?

On Seedance 2.5: up to 30 images (JPEG, PNG, WEBP or GIF, 20 MB each), 10 videos (MP4 or MOV, 200 MB each) and 10 audio files (WAV or MP3, 15 MB each), 50 files in total. Reference video and audio run 2-30 seconds each and 30 seconds combined. Use Omni Reference to mix material freely, or First & Last Frame to pin the opening and closing image of the shot.

Is there a free trial, and what does a video cost?

Signed-in accounts get a limited number of free runs on Seedance 2.0 Mini at 480p, up to 5 seconds. Everything else costs credits, billed by the second and scaling with the model and resolution you choose; if you attach a reference video, its length counts towards the billed seconds. The exact cost is shown next to the Generate button before you commit.

Can I upload a photo of a real person?

Not at the moment. The model provider does not currently accept realistic human faces in uploaded images or video, so a portrait reference will be rejected. Generating people from a text prompt works normally — the restriction is on what you upload, not on what the model draws.