Editly

Seedance 2.5 AI Video Generator

ByteDance's Seedance 2.5 generates a full 30-second video in a single shot — no stitching. Announced on 23 June 2026 and launched on 31 July, it takes up to 50 multimodal references, edits any region of a frame without regenerating the clip, and renders speech and lip-sync in step with the picture. Here is what actually shipped: confirmed specs, real pricing, and where you can run it today. Seedance 2.5 is coming soon to Editly.

Start creating

What Is Seedance 2.5? Features and Specs

ByteDance announced Seedance 2.5 on June 23, 2026 at the Volcano Engine FORCE conference and shipped it on July 31, 2026. Here is what changed from Seedance 2.0.

Native 30-Second Single Shot

Seedance 2.5 renders one continuous 30-second take in a single pass, doubling Seedance 2.0's roughly 15-second ceiling. No stitching, no characters drifting at the seams — the shot holds its subject, lighting, and camera path from first frame to last.

Up to 50 Multimodal References

Feed it as many as 30 reference images, 10 reference videos, and 10 audio clips in one request — up from 9 images, 3 videos, and 3 audio clips in Seedance 2.0. Characters, products, styles, and brand assets stay consistent across every scene you generate.

Region-Level Editing and Timestamp Control

Change part of a frame without regenerating the whole clip: remove a figure from the background, swap an object, or fix a single detail. Prompts can also assign actions to specific time ranges, so you direct what happens at second 4 and again at second 18.

Joint Audio and Multilingual Lip-Sync

Seedance 2.5 extends the unified audio-video architecture from Seedance 2.0, generating speech, ambience, and picture together. Dialogue lands in sync across languages, so a talking-head shot needs no separate lip-sync pass.

Production Control and Licensed IP

3D white-model previews, green-screen output, camera movement, and performance direction give teams shot-level control before the final render. ByteDance also ships licensed film IP creation with revenue sharing for authorized characters and stories.

Seedance 2.5 vs. Seedance 2.0: What Changed

Same unified audio-video foundation, a wider budget on nearly every axis. Reach for Seedance 2.5 when a cut needs 30 seconds of continuity, a pile of reference assets, or targeted edits — and stay on 2.0 when cost per second matters more.

Seedance 2.0
Seedance 2.52.5
Max single-shot durationUp to about 15 secondsUp to 30 seconds
Reference imagesUp to 9Up to 30
Reference videosUp to 3Up to 10
Reference audio clipsUp to 3Up to 10
Editing granularityRegenerate the whole clipRegion-level edits, rest of the frame untouched
DirectionWhole-clip promptingPer-timestamp action control
AudioJoint audio-video generationJoint audio-video plus multilingual lip-sync
Resolution through the API today480p / 720p480p / 720p
Best forShort cuts, fast iteration, lower cost per second30-second continuity, many references, targeted edits

Seedance 2.5 Pricing: Cost per Second

Aggregator API list prices in US dollars per second of output, with Seedance 2.0 alongside for reference. ByteDance's own Volcano Engine bills by token instead — roughly ¥70 per million billable tokens without video input and ¥42 with it, which works out to about ¥3.36 for a five-second 480p clip and ¥7.56 for a five-second 720p clip.

Seedance 2.0
Seedance 2.52.5
480p, per second of output$0.0775$0.140
720p, per second of output$0.165$0.315
720p with a video reference, per second$0.100$0.190
A 5-second 720p clipAbout $0.83About $1.58
The longest clip at 720p15 seconds, about $2.4830 seconds, about $9.45

How to Use Seedance 2.5: Step by Step

  1. 01

    Step 1: Pick Where to Run Seedance 2.5

    ByteDance's own apps came first — Jimeng (Dreamina) on the web and Doubao Pro — but both need a mainland Chinese phone number and a local payment method. Outside China you reach the model through Volcano Engine, the announced BytePlus ModelArk route, or a platform that resells API access.

  2. 02

    Step 2: Write the Prompt and Load References

    Describe the shot in plain language — subject, camera move, lighting, and what should happen when. Prompts run up to 30,000 characters, so a 30-second sequence can be directed beat by beat. Add reference images, video, or audio to lock a character, product, or style.

  3. 03

    Step 3: Set Duration, Resolution and Format

    Choose a length from 4 to 30 seconds, 480p or 720p, and an aspect ratio — 16:9, 9:16, 4:3, 3:4, 1:1, 21:9, or adaptive. Pin the framing with first and last frame images, leave audio generation on for synced speech, and export as MP4 or MOV.

Seedance 2.5 FAQ: Release Date, Pricing, 4K and API

Seedance 2.5 Is Coming Soon to Editly

While you wait, generate video on Seedance 1.5 Pro, Hailuo 2.3, Veo 3 and more — one credit balance, no Chinese phone number, no per-model subscription.

Try the AI Video Generator