Seedance 2.0 Mini Image-to-Video API

Overview

Seedance 2.0 Mini Image-to-Video is most useful when the difficult visual decisions have already been made in an image. The subject, composition, product design, color palette, and lighting begin with the uploaded asset, while the model is used to test how that visual should move.

This makes Mini different from a prompt-only generator. Instead of spending the request on inventing the entire scene, the Prompt can concentrate on motion: what the subject does, how the camera moves, what changes in the environment, and how quickly the shot develops.

The endpoint also separates three image-control jobs. A first_frame fixes the opening visual, an optional last_frame gives the shot a destination, and reference_image assets can guide appearance or visual direction without defining the exact first frame.

With 480P and 720P output and lower iCreat rates than the mainline Seedance 2.0 Image-to-Video route, Mini fits motion exploration, short-form variants, and batch review better than high-resolution final delivery.

Key Features

The source image carries the visual identity. The uploaded image establishes the subject design, composition, materials, lighting, and style before the model begins adding movement.

Three image roles solve different control problems. Use first_frame to lock the opening, last_frame to define the destination, and reference_image to guide appearance or environment without forcing an exact opening composition.

The Prompt describes change, not the still image. Strong I2V Prompts focus on action, camera movement, timing, environmental motion, and the details that must remain stable.

Low-cost motion testing. Mini makes it practical to generate several movement directions from the same asset, such as alternate camera moves, product reveals, gestures, or lighting transitions.

Reference video and audio can refine movement. Optional reference video can supply motion or camera rhythm, while reference audio can guide timing or sound character when a still image alone is not enough.

Audio can be created with the shot. generate_audio can add dialogue, ambience, sound effects, or music that develops with the animated image.

The resolution ceiling is part of the product choice. Mini stops at 720P. That keeps testing costs lower, but users who need 1080P or 4K should move selected concepts to the mainline endpoint.

Model Comparison

Seedance 2.0 Mini vs Mainline Image-to-Video

Seedance 2.0 Mini Image-to-Video vs Wan 2.7 Spicy vs HappyHorse 1.1 Spicy

Inputs

The Prompt and at least one image are required. Choose image roles according to the type of control needed.

Each image must have an aspect ratio from 0.4 to 2.5, dimensions from 300 to 6,000 pixels, and a file size under 30 MB. The total request body must remain under 64 MB.

Reference videos can be 2–15 seconds each, with a maximum combined duration of 15 seconds. Each file must remain under 200 MB. Reference audio files can also be 2–15 seconds each, with a maximum combined duration of 15 seconds and a maximum size of 15 MB per file.

Chinese Prompts can contain up to 2,000 characters, and English Prompts can contain up to 2,000 words. Set need_review: true when submitted media contains faces or copyrighted IP.

Parameters

Pricing

The examples below assume no reference video.

Total Cost = Unit Price × (Output Video Duration + Reference Video Duration)

Quick Start

This example uses a First Frame and an optional Last Frame to control a short product transition.

curl --fail-with-body --connect-timeout 10 --max-time 60 \
  -X POST https://api.icreat.ai/v1/task/submit/bytedance/seedance-2-0-mini/image-to-video \
  -H "Authorization: Bearer ${ICREAT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "content": [
      {
        "type": "text",
        "text": "Begin with the product in the supplied first frame. The camera moves slowly forward while soft light travels across the surface and the product rotates slightly. Keep the product shape, materials, colors, logo placement, and proportions consistent. Transition smoothly toward the supplied last frame. Add subtle mechanical ambience and restrained cinematic music."
      },
      {
        "type": "image_url",
        "image_url": {
          "url": "https://example.com/first-frame.jpg"
        },
        "role": "first_frame",
        "need_review": false
      },
      {
        "type": "image_url",
        "image_url": {
          "url": "https://example.com/last-frame.jpg"
        },
        "role": "last_frame",
        "need_review": false
      }
    ],
    "generate_audio": true,
    "ratio": "16:9",
    "resolution": "720p",
    "duration": 8,
    "watermark": false
  }'

The request returns a task_id.

Submit request → Receive task_id → Poll status → Retrieve result

Query the task status until it reaches SUCCEEDED, then use the same task_id to retrieve the result.

Output

Seedance 2.0 Mini Image-to-Video uses an asynchronous task flow.

The submission request returns a task_id. Task status is checked through the query-status endpoint, and the completed result is retrieved through the get-result endpoint after the task succeeds.

Detailed result fields are not listed until they are confirmed from an actual iCreat response.

Use Cases

Testing movement around one product hero image. Keep the same product composition while comparing a slow push-in, a partial rotation, a light sweep, a material reveal, or a faster social-media treatment.

Turning campaign key art into several short teasers. Use one approved character or scene image to test expression, clothing motion, background movement, camera direction, and audio without redesigning the visual for every version.

Building controlled before-and-after transitions. Supply a First Frame and Last Frame when a product, room, outfit, package, or lighting setup must move between two prepared states.

Animating a catalog without rebuilding every scene. Reuse existing product or category images and vary only the movement, pacing, aspect ratio, or sound treatment across a larger asset set.

Exploring motion from concept art. Start from a storyboard frame, environment painting, or character illustration to test whether the intended action and camera idea work before committing to a higher-resolution generation.

Limitations

Mini is limited to 480P and 720P. It is suitable for motion review and short-form output, but it cannot replace the 1080P or 4K delivery options of the mainline Seedance 2.0 Image-to-Video endpoint.

The source image is an anchor, not a pixel-locked frame. Faces, hands, product geometry, labels, textures, lighting, and small background elements may change once the model begins creating unseen motion.

Image roles can conflict when they ask for incompatible results. A First Frame, Last Frame, and several Reference Images should describe the same subject and a believable visual path. More references do not automatically produce stronger consistency.

Large rotations, rapid camera movement, major pose changes, or a Last Frame with a very different viewpoint increase the amount of visual information the model must invent. This raises the risk of drift, flicker, or unstable intermediate frames.

Output ratio also matters. A vertical or ultrawide ratio may require the model to crop, extend, or reposition parts of the uploaded image, even when the source composition is meant to stay stable.

Generated audio may contain unclear dialogue, distortion, or timing differences. Reference-video duration increases billed duration, and the endpoint does not expose a seed for exact repeatability.

FAQ

Should I use first_frame or reference_image?

Use first_frame when the uploaded image must be the exact opening composition. Use reference_image when the image should guide identity, appearance, environment, or style without fixing the first frame.

When does a last_frame improve the result?

A Last Frame helps when the shot must reach a prepared pose, product arrangement, lighting state, or final composition. It is less useful when the ending can develop freely.

Do more Reference Images always improve consistency?

No. Additional images help only when they provide compatible views or details. Conflicting faces, proportions, lighting, styles, or environments can make the result less stable.

Why can the subject still change when the source image is fixed?

The model must invent every view and movement that is not visible in the source. Large turns, hidden body parts, fast movement, and major camera changes create more opportunities for identity or detail drift.

When should I use Mini instead of Seedance 2.0 Image-to-Video?

Use Mini for lower-cost motion tests, batch variants, and 480P or 720P review outputs. Move selected concepts to the mainline endpoint when they need 1080P, 4K, or more demanding final-scene quality.