Turn Off Audio When You Do Not Need It
generate_audio is true by default, so every video comes with an audio track unless you pass generate_audio: false. Set it to false when you only need silent clips.
Input
first_frame_url
last_frame_url
Click to upload or drag and drop images.
Images: JPEG/JPG/PNG/WebP/GIF, Max 30MB
The text prompt or description for the video.
0/20000Click to upload or drag and drop
Supported formats: JPEG, PNG, WEBP, JPG, GIF Maximum file size: 30MB; Maximum files: 9
A list of input image URLs.
Click to upload or drag and drop
Supported formats: MP4, QUICKTIME Maximum file size: 50MB; Maximum files: 3
A list of input video URLs. Furthermore, the total length of the three videos must not exceed 15 seconds.
Click to upload or drag and drop
Supported formats: MPEG, WAV Maximum file size: 15MB; Maximum files: 3
A list of input audio URLs. Furthermore, the total length of the three audios must not exceed 15 seconds.
Whether to generate AI audio synchronized with the video.
The output video resolution.
The aspect ratio of the generated video.
Video duration in seconds.
Use online search
A configurable parameter. Defaults to true in the Playground.
There is no guarantee that everything can be filtered out; if you are not satisfied with the results, you will need to make your own arrangements.
Output
output typevideo
Examples
Explore different use cases and parameter configurations
README
Complete guide to using bytedance/seedance-2-mini
Seedance 2.0 Mini API: Text, Image and Reference to Video
Seedance 2.0 Mini is a ByteDance video generation model in the Seedance 2.0 series, offered on Kie.ai as bytedance/seedance-2-mini. It supports three input modes—text-to-video, image-to-video (first or first and last frames), and reference-to-video (images, videos, audio)—and outputs 480p or 720p video of 4 to 15 seconds, or a model-chosen length, with audio on by default.
Seedance 2.0 Mini API Parameters
Send requests to POST /api/v1/jobs/createTask with model set to bytedance/seedance-2-mini. Image-to-video (first frame), image-to-video (first and last frames) and multimodal reference-to-video are mutually exclusive and cannot be combined in the same request.
| Parameter | Details |
|---|---|
| model | Required. Set to bytedance/seedance-2-mini. |
| prompt | Required. Text prompt from 3 to 20000 characters. |
| first_frame_url | Optional. Image URL or asset://{assetId} for the opening frame in image-to-video mode. Cannot be combined with any reference_* field. |
| last_frame_url | Optional. Image URL or asset://{assetId} for the closing frame when used with first_frame_url. Cannot be combined with any reference_* field. |
| reference_image_urls | Optional. Up to 9 images in jpeg, png, webp, or gif format. Aspect ratio (width/height) must be between 0.4 and 2.5, width and height between 300 and 6000 px, and file size under 30 MB each. Cannot be combined with first_frame_url or last_frame_url. |
| reference_video_urls | Optional. Up to 3 videos in mp4 or mov format at 480p or 720p. Each video must be 2 to 15 seconds, with a combined total of at most 15 seconds. Aspect ratio 0.4 to 2.5, width and height 300 to 6000 px, total pixels 409,600 to 927,408, 24 to 60 FPS, and up to 50 MB each. |
| reference_audio_urls | Optional. Up to 3 audio clips in wav or mp3 format. Each clip must be 2 to 15 seconds, with a combined total of at most 15 seconds, and up to 15 MB each. |
| generate_audio | Optional boolean. Default is true, which generates native synchronized audio. |
| resolution | Optional. 480p (which the documentation describes as "for faster generation") or 720p (described as "for balance"). Default is 720p. |
| aspect_ratio | Optional. Accepted values: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, or adaptive. Default is 16:9. |
| duration | Optional. Video length from 4 to 15 seconds, or -1 to let the model choose the duration. Default is 5. |
| web_search | Optional boolean. Enables online search. Supported in text-to-video mode only. |
What Seedance 2.0 Mini Generated in Our Tests
Each video was generated with Seedance 2.0 Mini on Kie.ai in October 2026. We checked that each output has an audio track but did not judge the audio itself.
Text to video: a rainy cafe window
We asked for a rainy evening outside a small cafe window with raindrops running down the glass, warm light inside, a person in a yellow raincoat walking past on the sidewalk, a slow camera push-in toward the window, and sounds of steady rain, distant traffic, and footsteps on wet pavement, at 720p, 16:9, and 5 seconds. The API returned a 1280x720 MP4 of 5.09 seconds with an audio track. The output shows rain on the glass, warm interior light, and a person in a yellow raincoat, but the camera moves through the window out into the street instead of pushing in toward the window as asked.
First frame: the robot waves
Starting from a square first-frame image of our orange robot mascot on a plain light-grey studio background, we asked for the robot to wave at the camera with its right hand, tilt its head slightly to the side, and blink its antenna lights while keeping the same design and colors, at 480p, adaptive aspect ratio, and 5 seconds. We got back a 640x640 MP4 of 5.09 seconds with an audio track. The adaptive setting kept the square shape of the input image, and the robot kept its orange color, blue eyes, and lightning-bolt chest badge and waved, but it also walked toward the camera, which the prompt did not ask for, and the head tilt is not visible.
First and last frames: studio to valley
Using the studio robot image as the first frame and an illustrated image of the robot standing on a grassy hill in a sunny valley with a river and a hot-air balloon as the last frame, we asked for a smooth camera move carrying the view from the robot in the studio out to the robot standing in the sunny valley landscape, at 480p, 16:9, and 5 seconds. The API returned an 864x496 MP4 of 5.09 seconds with an audio track. The video starts on the studio robot and ends on the valley scene of the last frame with the robot waving on the hill, but the change from studio to valley happens as a cut between about 1.5 seconds and 2.0 seconds rather than the smooth camera move requested.
Two reference images, duration -1
Passing two reference images—the studio robot image first and the valley image showing the robot in the landscape second—we asked for the robot from the first reference image to walk across the grassy valley from the second reference image toward the river, stop, and wave, with cheerful background music, at 720p, 16:9, and duration -1. The API returned a 1280x720 MP4 of 10.08 seconds with an audio track, as duration -1 let the model choose the length. In the video, the robot walks across the grassy hillside toward the river, stops at the water's edge, and waves, matching the design and colors of the reference images.
Seedance 2.0 Series Comparison on Kie.ai
The table below compares only the documented parameters of the three Seedance 2.0 models available on Kie.ai.
| Seedance 2.0 Mini | Seedance 2.0 | Seedance 2.0 Fast | |
|---|---|---|---|
| Model ID | bytedance/seedance-2-mini | bytedance/seedance-2 | bytedance/seedance-2-fast |
| Resolution | 480p, 720p | 480p, 720p, 1080p | 480p, 720p |
| Duration | 4-15 s or -1 | 4-15 s or -1 | 4-15 s or -1 |
| Reference inputs | Up to 9 images, 3 videos, 3 audio clips | Up to 9 images, 3 videos, 3 audio clips | Up to 9 images, 3 videos, 3 audio clips |
| return_last_frame | Not in documentation | Supported | Supported |
| web_search | Text-to-video only | Text-to-video only | Text-to-video only |
How to Use the Seedance 2.0 Mini API on Kie.ai
1. Create an API Key and Authenticate
2. Submit a Task to createTask
3. Retrieve the Generated Video
Tips Before You Ship
Check Aspect Ratios on Reference and Last-Frame Images
The documentation specifies an aspect ratio (width/height) limit of 0.4 to 2.5 for reference_image_urls. In our October 2026 tests on Kie.ai, passing a wide banner image with an aspect ratio of 4.03 as last_frame_url also failed within about a second with the error "expected the aspect ratio to be between 0.40 and 2.50, but received image with aspect ratio: 4.03" (consuming 0 credits). Keep last-frame images within 0.4 to 2.5 as well.
Use First and Last Frame Mode for Exact Frame Matching
In multimodal reference mode, you can name a reference image as the first or last frame in your text prompt. However, according to the Kie.ai documentation, to strictly guarantee that the opening and closing frames match your provided images, use image-to-video (first and last frames) mode instead.
Account for Reference Video and Auto-Duration Billing
When your request includes reference videos, billing uses a lower per-second price multiplied by the total reference-video duration plus the output duration. When duration is set to -1, billing follows the duration actually generated.
Why Use Seedance 2.0 Mini on Kie.ai
Unified Authentication and Credit Balance
All models in the Kie.ai market use the same authentication method: an Authorization: Bearer YOUR_API_KEY header with the key from your Kie.ai API Key Management page. Your account has one credit balance that all generation services draw from; check it with the Get Remaining Credits API.
Browser Playground and API in One Place
Kie.ai provides this model as an API with a playground; there is no app to download or install. You can try this model in the playground on this page without writing code.
Free Credits for Initial Testing
New Kie.ai users receive 80 free credits for testing.
New Kie.ai users receive 80 free credits for testing. After that, generations draw from your account's credit balance; prices are shown in the pricing card above. You can try this model in the playground on this page without writing code. Kie.ai provides this model as an API with a playground; there is no app to download or install.
Prices are shown in the pricing card above. When you use reference videos, the documentation states that billing uses a lower per-second price multiplied by the sum of the total reference-video duration and the output duration. Your account has one credit balance that all generation services draw from; check it with the Get Remaining Credits API.
Seedance 2.0 Mini is a ByteDance video generation model in the Seedance 2.0 series, offered on Kie.ai under the model ID bytedance/seedance-2-mini. It generates 480p or 720p videos of 4 to 15 seconds, or a model-chosen length with duration -1, from text, first/last frames, or multimodal references (up to 9 images, 3 videos, and 3 audio clips). Synchronized audio is on by default. On its Lumina site, ByteDance calls Seedance 2.0 Mini a cost-efficient option for frequent publishing and rapid concept testing.
In the documented parameters on Kie.ai, Seedance 2.0 Mini (bytedance/seedance-2-mini) outputs 480p or 720p and does not list a return_last_frame parameter, whereas Seedance 2.0 (bytedance/seedance-2) outputs 480p, 720p, or 1080p and includes return_last_frame. Both models support durations of 4 to 15 seconds or -1, up to 9 reference images, 3 reference videos, and 3 reference audio clips, and web_search in text-to-video mode. See the parameter table above for details, and check the pricing cards for each model's cost.
In the documented parameters on Kie.ai, both Seedance 2.0 Mini (bytedance/seedance-2-mini) and Seedance 2.0 Fast (bytedance/seedance-2-fast) support 480p and 720p resolutions, durations of 4 to 15 seconds or -1, the same reference input limits (up to 9 images, 3 videos, 3 audio clips), and web_search in text-to-video mode. The documented parameter difference is that Seedance 2.0 Fast includes return_last_frame, while Seedance 2.0 Mini does not list return_last_frame in its documentation. See the parameter table above for full parameter details, and check the pricing cards for each model's cost.
Create a key on the API Key Management page and pass it in the request header when calling POST /api/v1/jobs/createTask. All models in the Kie.ai market use the same authentication method: an Authorization: Bearer YOUR_API_KEY header with the key from your Kie.ai API Key Management page.
No. Image-to-video (first frame), image-to-video (first and last frames), and multimodal reference-to-video are mutually exclusive and cannot be used in the same request; do not combine first_frame_url or last_frame_url with any reference_* field. In reference mode, a reference image can be named as the first or last frame in the prompt, but to strictly guarantee the first and last frames match the given images, use image-to-video (first and last frames). See the parameter table above for field rules.
Yes. Native synchronized audio is enabled by default via the generate_audio parameter (boolean, default true). In multimodal reference-to-video mode, you can also pass up to 3 wav or mp3 clips (2 to 15 seconds each, up to 15 seconds total, up to 15 MB each) in reference_audio_urls. See the parameter table above for specifications.
Setting duration to -1 lets the model choose the output video length instead of fixing it to a number between 4 and 15 seconds. For example, in our October 2026 test on Kie.ai with two reference images and duration set to -1, the model returned a 10.08-second MP4. When duration is -1, billing follows the duration actually generated. See the parameter table above for all duration options.