Flint Image 4K is here. Frontier quality, at a fraction of the price.

Seedance 2.5 API

Video·seedance-2.5·by Bytedance

Seedance 2.5 Image-to-Video animates first or first-and-last frames into cinematic clips up to 30 seconds with synchronized audio and natural motion.

Try it now →
API reference
Runtime (p50)
4m
Estimated price
From $1.15
Call the API
prediction.sh
sh
curl -X POST \
  -H "Authorization: Bearer $EACHLABS_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "bytedance-seedance-2-5-image-to-video",
    "version": "0.0.1",
    "input": {
        "prompt": "The camera pushes in rapidly toward the bright mirrored reflection on the left side of the cat's iris, not toward the black pupil. The reflected desert horizon inside the reflection grows and expands until it fills the entire frame, becoming the real landscape at golden hour. The camera then drifts slowly forward and downward over the dune ridges, sand streaming past the lens, gradually revealing the lone figure walking along the distant crest.",
        "duration": "6",
        "image_url": "https://cdn-us.eachlabs.ai/defaults/58a5b6e7e3bb4918a49cf711441066c9.png",
        "resolution": "720p",
        "end_image_url": "https://cdn-us.eachlabs.ai/defaults/220008cd852242b684ea10e249ec5813.png",
        "output_format": "mp4",
        "generate_audio": true
    },
    "webhook_url": ""
}' \
  https://api.eachlabs.ai/v1/prediction/
Documentation8 sections
  • Overview

    Bytedance | Seedance 2.5 | Image to Video Overview

    Bytedance | Seedance 2.5 | Image to Video turns a still image, or a first-and-last frame setup, into a cinematic clip with natural motion and synchronized audio. It is part of ByteDance’s Seedance 2.5 family, which is designed for controlled AI video generation across image, video, and audio inputs.

    Its main differentiator is native long-form generation: Seedance 2.5 supports a 30-second single take and can extend into longer sequences through multi-round generation, while keeping motion and scene continuity stable. On each::labs, this makes the model useful for creators who want a strong image-to-video workflow without stitching together many short clips.

    Compared with earlier Seedance releases, the 2.5 family also emphasizes richer reference control, region-level editing, and synchronized audiovisual output.

  • Capabilities

    Capabilities

    • Animates a still image into a cinematic video clip with natural motion.
    • Supports 30-second single-pass generation in the Seedance 2.5 family.
    • Can extend sequences beyond one shot through multi-round continuation.
    • Produces synchronized audio in the broader Seedance 2.5 system.
    • Supports controlled motion direction, including camera movement and scene pacing.
    • Works with reference-guided workflows that help preserve style and subject consistency.
    • Allows more precise editing modes in the family, including region-level changes and timestamp-based control.
    • Supports high-resolution output, with secondary reporting up to 4K.
  • Use cases

    Use Cases for Bytedance | Seedance 2.5 | Image to Video

    Creators can turn a single illustration or still frame into a polished social clip. For example: “Animate this character portrait with a slow push-in, blinking eyes, and subtle hair movement.” This is a strong fit because the model is built for natural motion from image input.

    Marketers can create short product animations from a hero image. Example: “Clean studio shot, gentle turntable motion, reflective highlights, premium commercial look.” The model’s cinematic motion and high-resolution output are useful when the image already captures the product clearly.

    Designers can preview concept art as moving art direction references. Example: “City skyline at dusk, light fog drifting, slight camera pan, calm mood.” Seedance 2.5’s controlled motion and longer clip length help preserve the design intent across more time.

    Developers building on the Bytedance | Seedance 2.5 | Image to Video API can prototype image-led motion experiences for apps, galleries, and campaigns where a still asset needs a realistic animated reveal.

  • Tips & tricks

    Tips and Tricks

    For the best results with Bytedance | Seedance 2.5 | Image to Video, describe motion in simple, film-like language. State what should move, what should stay stable, and how the camera should behave. Seedance 2.5 coverage highlights camera-path control, region-level editing, and reference-guided generation, so prompts that name the subject, motion, and framing usually perform better than vague style prompts.

    Use short, specific instructions such as timing, atmosphere, and shot direction. If your scene has multiple elements, keep the main action singular and explicit. Example prompts: “Slow dolly-in on a lone cyclist at sunset, subtle wind in the trees, realistic shadows.” “Static camera, product packaging rotates gently, clean studio lighting, premium commercial look.” “Portrait subject blinks, hair moves softly, cinematic background parallax.”

    When using the Bytedance | Seedance 2.5 | Image to Video API, test one variable at a time: motion first, then camera movement, then lighting or atmosphere. That workflow makes it easier to isolate what the model is responding to.

  • Technical spec

    Technical Specifications

    • Input type: Image-to-video generation from a source image, with prompt guidance; the Seedance 2.5 family also supports first-and-last frame workflows and broader multimodal references.
    • Output duration: Up to 30 seconds in a single generation, with multi-round extension for longer sequences.
    • Resolution: Reported support includes up to 4K, with 1080p and lower-resolution outputs also referenced in secondary coverage.
    • Aspect ratios: Common supported ratios in Seedance workflows include 16:9, 9:16, 4:3, 3:4, 21:9, and 1:1.
    • Input formats: Images plus text prompts; the wider Seedance 2.5 system also accepts image, video, and audio reference assets.
    • Output format: Video output is typically delivered as an MP4-style clip in public model descriptions.
    • Processing time: No official average runtime was confirmed in the sources provided; generation time is expected to vary with resolution, duration, and reference complexity.
  • Things to be aware of

    Things to Be Aware Of

    Seedance 2.5 performs best when the source image is sharp and the subject is visually clear. Complex scenes with many moving parts can reduce consistency, especially if the prompt asks for several actions at once.

    Users often get weaker results when they overload the prompt with conflicting motion cues or ask for large changes that are not present in the input image. Long clips also increase the chance of drift, so continuity should be checked carefully when extending beyond a single generation.

    Resource demand may rise with longer duration, higher resolution, and more detailed reference control, so practical workflows usually start with a simple motion test before moving to final output.

  • Key considerations

    Key Considerations

    Bytedance | Seedance 2.5 | Image to Video is best when the source image already contains a clear subject, composition, and visual style. The model is positioned for cinematic motion, so it works especially well for product shots, character portraits, and stylized scenes that need graceful movement rather than chaotic action.

    Users should plan for prompt-driven control: the more clearly you describe camera motion, subject behavior, and scene tone, the more consistent the result is likely to be. Seedance 2.5 also appears optimized for high-control workflows, which can make it more suitable than generic video generators when you need continuity across longer clips or more precise scene direction.

  • Limitations

    Limitations

    Bytedance | Seedance 2.5 | Image to Video cannot reliably invent every missing detail from weak source images, and it is not ideal for scenes that require exact frame-perfect replication over many seconds.

    Public sources in this research did not confirm a guaranteed average processing time, so latency should be treated as variable. Some capability details also differ across secondary reports, especially around resolution and reference limits, so production use should rely on the exact each::labs configuration exposed for this model.

Related models

4 models
PixVerse V6 Transition AI model preview
PixVerse V6 TransitionPixverse
FFmpeg API · Images to Video AI model preview
FFmpeg API · Images to VideoFfmpeg Api
MiniMax H3 · Image-to-Video AI model preview
MiniMax H3 · Image-to-VideoMinimax
PixVerse C1 Transition AI model preview
PixVerse C1 TransitionPixverse
* FAQ

About Seedance 2.5 API

01 / 05

What is Seedance 2.5 Image-to-Video?

Seedance 2.5 Image-to-Video is ByteDance's video generation model that animates a still image into a moving clip. Your image becomes the first frame, and the model builds motion, camera work, and synchronized audio around it, producing up to 30 seconds of video that stays true to the original composition.