ByteDance

Seedance 2

Seedance 2.5 combines text-only generation with optional image, video, and audio guidance in one flagship task. It supports 480p and 720p output from 4 to 15 seconds, with up to nine images, three videos, and three audios. Seedance 2.0 remains available in Full, Fast, and Mini tiers; each has an LR counterpart with more permissive prompt filtering when avoiding false refusals matters.

How to use this model

1
Choose 2.5 for the flagship workflow; use a Seedance 2.0 LR version when prompt acceptance matters or a standard 2.0 variant rejected a harmless request.
2
Assign each reference a clear job in the prompt—for example character, location, camera movement, motion rhythm, or soundtrack.
3
For first-and-last-frame generation, upload both frames and describe the action that makes the transition plausible.
4
Include dialogue, ambience, effects, and music explicitly when native audio is part of the result.
5
Use Fast or Mini for iteration, then compare the chosen version on the same prompt and references before finalizing.

2.5Available

2.5

Seedance 2.5 uses one multimodal workflow for text-only and reference-guided generation. Add up to nine images, three videos, and three audio references, then address them by position in the prompt. Output is available at 480p or 720p for 4–15 seconds; a supplied image determines the aspect ratio.

Try now

Properties

Single task type for text and mixed image, video, and audio guidance.
Supports up to nine images, three videos, and three audio references.

Best for

High-quality text-to-video generation
Image-guided subject and style control
Mixed image, video, and audio references

Avoid for

1080p output, output longer than 15 seconds, or audio-only reference generation

Tips

Add only the image, video, and audio references needed for the result.
Use @imageN, @videoN, and @audioN in the prompt to address references by position.

How to use this model

1
Start with a clear ordered description of the scene, motion, camera, and audio.
2
Attach only the image, video, and audio references needed for the result.
3
Use @imageN, @videoN, and @audioN to identify each reference in the prompt.
4
Choose 480p or 720p and a duration from 4 to 15 seconds.

Parameters

Prompt

prompt

Required

Text prompt describing the desired scene, motion, and action.

Resolution

resolution

Optional

Output video resolution. This tier does not support 1080p.

Default: 720p

480p
720p

Duration

duration

Optional

Generated video duration in seconds.

Default: 5Range: 4 – 15

Aspect Ratio

aspect_ratio

Optional

Output aspect ratio. A reference image takes precedence when supplied.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
auto

Reference Images

image_urls

Optional

Up to nine images for subject appearance, content, or style guidance.

Reference Videos

video_urls

Optional

Up to three motion, content, or style references totaling at most 15.4 seconds.

Reference Audio

audio_urls

Optional

Up to three MP3, OGG, WAV, M4A, or AAC references; audio requires an image or video reference.

2.0Available

2.0

This full Seedance 2.0 workflow can generate from text, animate a first frame, interpolate to a last frame, or use several reference media types. References are mutually structured inputs: use the dedicated first-frame fields for interpolation and the reference collections for multimodal guidance.

Try now

Properties

Inputs: prompt, optional first/last frames, and image, video, or audio references.
Controls include duration, resolution, aspect ratio, seed, and native audio.

How to use this model

1
Choose text, first-frame, first/last-frame, or multimodal-reference generation.
2
Assign each image, video, or audio reference a specific role in the prompt.
3
Set duration, resolution, aspect ratio, and native-audio generation together.
4
Review subject consistency, reference adherence, motion, and sound synchronization.

Examples

Image-to-video material motion generated with Seedance 2.

A reviewed video output created with 2.0 - Text to Video in Omnipix.

Parameters

Seed

seed

Optional

Random seed. Set for reproducible generation.

Image

image

Optional

Input image for image-to-video generation (first frame). Cannot be combined with reference images.

Prompt

prompt

Required

Text prompt for video generation. Maximum 4000 characters. For best results, keep prompts under 600 English words.

Duration

duration

Optional

Video duration in seconds. Automatic duration is unavailable because the generation is quoted and reserved before processing completes.

Default: 5Range: -1 – 15

Resolution

resolution

Optional

Video resolution. 4K outputs 10-bit H.265/HEVC at high bitrate.

Default: 720p

480p
720p
1080p
4k

Aspect Ratio

aspect_ratio

Optional

Video aspect ratio. Set to adaptive to let the model choose the best ratio based on inputs.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
9:21
adaptive

Generate Audio

generate_audio

Optional

Generate synchronized audio with the video, including dialogue, sound effects, and background music.

Default: true

Last Frame Image

last_frame_image

Optional

Input image for last frame generation. Only works if a first frame image is also provided. Cannot be combined with reference images.

Reference Audios

reference_audios

Optional

Reference audio files for audio-driven generation and lip-sync. Requires at least one reference image or video. Reference them in your prompt as [Audio1], [Audio2], etc.

Reference Images

reference_images

Optional

Reference images for character consistency, style guidance, and scene composition. Cannot be used together with first/last frame images.

Reference Videos

reference_videos

Optional

Reference videos for motion transfer, style reference, and editing. Reference them in your prompt as [Video1], [Video2], etc.

Recommended2.0Available

2.0 LR

LR means Less Restrictive and is the recommended full-quality Seedance 2.0 choice when prompt acceptance matters. It substantially reduces moderation false positives that can block benign prompts or synthetic people, while keeping the complete text, frame, image, video, and audio workflow. Filtering is permissive up to NSFW-level requests; platform rules still apply, so acceptance is not guaranteed. Omnipix infers the generation mode from the attached inputs.

Start here when ordinary Seedance rejects harmless prompts. Full LR keeps full quality with substantially more permissive filtering—even for synthetic people and requests up to NSFW level—subject to platform rules.

Try now

Properties

Less-restrictive prompt filtering, including prompts up to NSFW level.
Input-driven text, frame-guided, and multimodal-reference workflows.

Best for

Text-to-video generation
First and last frame control
Mixed image, video, and audio references

Avoid for

Output longer than 15 seconds

Tips

Omnipix infers text-to-video, frame-guided, or multimodal-reference generation from the supplied inputs.
Use @imageN, @videoN, and @audioN in the prompt to address references by position.

How to use this model

1
Add no media for text-to-video, one or two frames for frame guidance, or image, video, and audio references for multimodal guidance.
2
Use LR deliberately when the prompt needs less-restrictive filtering, including NSFW-level content.
3
Describe how each supplied source affects the result.
4
Set output controls and inspect reference adherence after generation.

Examples

A reviewed video output created with 2.0 LR in Omnipix.

Parameters

Prompt

prompt

Required

Text prompt describing the desired scene, motion, and action.

Resolution

resolution

Optional

Output video resolution.

Default: 480p

480p
720p
1080p

Duration

duration

Optional

Generated video duration in seconds.

Default: 5Range: 4 – 15

Aspect Ratio

aspect_ratio

Optional

Output aspect ratio. Auto lets the model derive the ratio from the supplied frames.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
auto

Reference Images

image_urls

Optional

One or two images can define first and last frames; larger image sets provide multimodal guidance.

Reference Videos

video_urls

Optional

Optional motion or style references for multimodal guidance.

Reference Audio

audio_urls

Optional

Optional MP3 or WAV soundtrack references; audio requires an image or video reference.

2.0 FastAvailable

2.0 Fast

Use Fast for shorter iteration cycles while retaining text-to-video, image-to-video, first/last-frame, and multimodal-reference inputs. Because it exposes the same source fields as the full version, the same prompt and references can be compared directly.

Try now

Properties

Fast variant with text, image, video, and audio reference inputs.
Supports first/last-frame guidance and optional generated audio.

How to use this model

1
Build the prompt and reference set exactly as for the full Seedance 2.0 version.
2
Use first/last-frame fields for a fixed transition, not the general reference collection.
3
Set output and audio controls before running the iteration.
4
Compare Fast and full outputs on identical inputs when selecting a final render.

Parameters

Seed

seed

Optional

Random seed. Set for reproducible generation.

Image

image

Optional

Input image for image-to-video generation (first frame). Cannot be combined with reference images.

Prompt

prompt

Required

Text prompt for video generation. Maximum 4000 characters. For best results, keep prompts under 600 English words.

Duration

duration

Optional

Video duration in seconds. Automatic duration is unavailable because the generation is quoted and reserved before processing completes.

Default: 5Range: -1 – 15

Resolution

resolution

Optional

Video resolution.

Default: 720p

480p
720p

Aspect Ratio

aspect_ratio

Optional

Video aspect ratio. Set to adaptive to let the model choose the best ratio based on inputs.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
9:21
adaptive

Generate Audio

generate_audio

Optional

Generate synchronized audio with the video, including dialogue, sound effects, and background music.

Default: true

Last Frame Image

last_frame_image

Optional

Input image for last frame generation. Only works if a first frame image is also provided. Cannot be combined with reference images.

Reference Audios

reference_audios

Optional

Reference audio files for audio-driven generation and lip-sync. Requires at least one reference image or video. Reference them in your prompt as [Audio1], [Audio2], etc.

Reference Images

reference_images

Optional

Reference images for character consistency, style guidance, and scene composition. Cannot be used together with first/last frame images.

Reference Videos

reference_videos

Optional

Reference videos for motion transfer, style reference, and editing. Reference them in your prompt as [Video1], [Video2], etc.

Recommended2.0Available

2.0 Fast LR

Fast LR is the recommended iteration variant when standard Seedance rejects benign prompts or synthetic people. It keeps the image, video, and audio reference workflow of Seedance 2.0 Fast while substantially reducing moderation false positives. Prompts may reach NSFW level subject to platform rules, and acceptance is not guaranteed. The workflow is inferred from the inputs.

Choose Fast LR for quicker iterations with substantially fewer false prompt refusals, including benign synthetic-person scenes. Requests up to NSFW level are supported subject to platform rules.

Try now

Properties

Fast generation with input-driven multimodal workflows.
Less-restrictive filtering, including prompts up to NSFW level.

Best for

Text-to-video generation
First and last frame control
Mixed image, video, and audio references

Avoid for

1080p output or output longer than 15 seconds

Tips

Omnipix infers text-to-video, frame-guided, or multimodal-reference generation from the supplied inputs.
Use @imageN, @videoN, and @audioN in the prompt to address references by position.

How to use this model

1
Supply only the media needed for the intended workflow; Omnipix infers the mode from those inputs.
2
Add only the necessary URLs and identify them in the prompt.
3
Set output duration, resolution, and aspect ratio.
4
Review reference use, motion, and audio together.

Parameters

Prompt

prompt

Required

Text prompt describing the desired scene, motion, and action.

Resolution

resolution

Optional

Output video resolution. This tier does not support 1080p.

Default: 480p

480p
720p

Duration

duration

Optional

Generated video duration in seconds.

Default: 5Range: 4 – 15

Aspect Ratio

aspect_ratio

Optional

Output aspect ratio. Auto lets the model derive the ratio from the supplied frames.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
auto

Reference Images

image_urls

Optional

One or two images can define first and last frames; larger image sets provide multimodal guidance.

Reference Videos

video_urls

Optional

Optional motion or style references for multimodal guidance.

Reference Audio

audio_urls

Optional

Optional MP3 or WAV soundtrack references; audio requires an image or video reference.

2.0Available

2.0 Mini

Mini exposes the same basic text and reference structure in a smaller Seedance variant. Omnipix infers the workflow from the supplied media. Use it for early exploration and keep prompts and sources organized so a promising setup can be moved unchanged to another version.

Try now

Properties

Mini variant with prompt and image, video, and audio inputs.
Automatically infers the workflow from supplied inputs.

Best for

Text-to-video generation
First and last frame control
Mixed image, video, and audio references

Avoid for

1080p output or output longer than 15 seconds

Tips

Omnipix infers text-to-video, frame-guided, or multimodal-reference generation from the supplied inputs.
Use @imageN, @videoN, and @audioN in the prompt to address references by position.

How to use this model

1
Assemble the minimum required references; the workflow is inferred from what you attach.
2
Describe each source's role and the desired shot progression.
3
Set duration, resolution, and aspect ratio before generating.
4
Save the exact prompt and inputs when moving a successful draft to another version.

Parameters

Prompt

prompt

Required

Text prompt describing the desired scene, motion, and action.

Resolution

resolution

Optional

Output video resolution. This tier does not support 1080p.

Default: 720p

480p
720p

Duration

duration

Optional

Generated video duration in seconds.

Default: 5Range: 4 – 15

Aspect Ratio

aspect_ratio

Optional

Output aspect ratio. Auto lets the model derive the ratio from the supplied frames.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
auto

Reference Images

image_urls

Optional

One or two images can define first and last frames; larger image sets provide multimodal guidance.

Reference Videos

video_urls

Optional

Optional motion or style references for multimodal guidance.

Reference Audio

audio_urls

Optional

Optional MP3 or WAV soundtrack references; audio requires an image or video reference.

Recommended2.0Available

2.0 Mini LR

Mini LR is the recommended economical starting point when ordinary Seedance rejects benign prompts or synthetic people. It combines the compact, input-driven Seedance workflow with substantially more permissive filtering and accepts requests up to NSFW level subject to platform rules. Acceptance is not guaranteed.

Choose Mini LR for lower-cost exploration with substantially fewer false prompt refusals, including benign synthetic-person scenes. Requests up to NSFW level are supported subject to platform rules.

Try now

Properties

Compact generation with input-driven multimodal workflows.
Less-restrictive filtering, including prompts up to NSFW level.

Best for

Text-to-video generation
First and last frame control
Mixed image, video, and audio references

Avoid for

1080p output or output longer than 15 seconds

Tips

Omnipix infers text-to-video, frame-guided, or multimodal-reference generation from the supplied inputs.
Use @imageN, @videoN, and @audioN in the prompt to address references by position.

How to use this model

1
Supply only the sources needed for the task; Omnipix infers the mode automatically.
2
Attach only relevant image, video, and audio sources.
3
Make every reference role explicit in the prompt.
4
Set output controls and inspect the result against the supplied sources.

Parameters

Prompt

prompt

Required

Text prompt describing the desired scene, motion, and action.

Resolution

resolution

Optional

Output video resolution. This tier does not support 1080p.

Default: 720p

480p
720p

Duration

duration

Optional

Generated video duration in seconds.

Default: 5Range: 4 – 15

Aspect Ratio

aspect_ratio

Optional

Output aspect ratio. Auto lets the model derive the ratio from the supplied frames.

Default: 16:9

21:9
16:9
4:3
1:1
3:4
9:16
auto

Reference Images

image_urls

Optional

One or two images can define first and last frames; larger image sets provide multimodal guidance.

Reference Videos

video_urls

Optional

Optional motion or style references for multimodal guidance.

Reference Audio

audio_urls

Optional

Optional MP3 or WAV soundtrack references; audio requires an image or video reference.