Yapper Docs
Models & Inputs

Video generation

Video models with per-length pricing and input fields.

POST /api/v1/processes with "type": "video-generation". Pick any model — the API does not plan-gate models. GET /api/v1/models returns the same capabilities at runtime.

Seedance generated audio

Seedance 2.x generates a native audio track by default. Set generateAudio: false for a silent video. When referenceAudios are attached, generated audio is required and generateAudio must be true or omitted.

{
  "type": "video-generation",
  "model": "seedance-2.5",
  "input": {
    "prompt": "A silent cinematic landscape reveal",
    "videoLength": 5,
    "generateAudio": false
  }
}

Genjutsu workflow

Use model: "genjutsu" with exactly one source video of 4 to 30 seconds in referenceVideos. Set genjutsuMode to "motion-transfer" (the default) to rebuild the scene using the source motion, or "object-swap" to replace a specified element. Add 1 to 8 referenceImages. Output can be 480p, 720p, or 1080p; duration and aspect ratio follow the source video.

model: "genjutsu-light" takes the same input at a lower price, for source videos of 4 to 15 seconds.

{
  "type": "video-generation",
  "model": "genjutsu",
  "input": {
    "prompt": "Replace the jacket with the one in the reference image",
    "genjutsuMode": "object-swap",
    "referenceVideos": [{ "assetId": "source_video_asset_id" }],
    "referenceImages": [{ "assetId": "jacket_image_asset_id" }],
    "resolution": 720
  }
}

Seedance 2.5 Edit workflow

Use model: "seedance-2.5-edit" for the dedicated base-video workflow. Put the base video first in referenceVideos; later video items are additional references. aspectRatio and durationMode are always "auto", and videoLength is not accepted.

editOperation defaults to "edit". Edit requests require a non-empty prompt; "extend", "sequel", and "prequel" work without one because the server supplies the complete continuation/lead-in instruction.

{
  "type": "video-generation",
  "model": "seedance-2.5-edit",
  "input": {
    "editOperation": "sequel",
    "referenceVideos": [{ "assetId": "base_video_asset_id" }]
  }
}

Seedance 2.5 edit retries

With a referenceVideos item, Seedance 2.5 can read the prompt as an edit of that video. An edit keeps the reference video's length and aspect ratio, so when the request asks for a different videoLength or aspectRatio, the process fails and says so.

Set retryAsEdit: true to retry it once as an edit instead, with durationMode and aspectRatio on "auto". The output then matches the reference video. The retry is charged for the reference video's length before it starts, so a team that can't cover that fails as out of credits. Any difference from the delivered length is settled when it completes. Use model: "seedance-2.5-edit" when you want an edit from the start.

{
  "type": "video-generation",
  "model": "seedance-2.5",
  "input": {
    "prompt": "Recreate this clip at dusk",
    "videoLength": 10,
    "aspectRatio": "16:9",
    "referenceVideos": [{ "assetId": "reference_video_asset_id" }],
    "retryAsEdit": true
  }
}

Input fields

FieldTypeRequiredDescription
promptstringyesWhat to generate.
genjutsuMode"motion-transfer" | "object-swap"noOnly for genjutsu and genjutsu-light. Defaults to "motion-transfer"; use "object-swap" to replace a specific element in the source video.
editOperation"edit" | "extend" | "sequel" | "prequel"noOnly for seedance-2.5-edit. Defaults to "edit"; prompt is optional for "extend", "sequel", and "prequel".
aspectRatiostringnoOne of the model's aspect ratios, e.g. "16:9". Use "adaptive" when advertised to follow the primary image/video reference.
resolutionnumbernoOutput resolution height, e.g. 1080.
draftbooleannoOnly for seedance-2.5 and seedance-2.5-edit. Every 480p run is a draft by default, billed like any 480p run; send false for a plain 480p render that cannot be finished. true sets resolution to 480, which must then stay 480. A completed draft reports draft.finishableUntil and draft.finishResolutions. Then finish it on seedance-2.5 with draftVideoId and resolution 1080 (the default) or 2160 (4K), as listed in its draft.finishResolutions, edit drafts included.
draftVideoIdstringnoOnly for seedance-2.5. The asset id of a completed draft from seedance-2.5 or seedance-2.5-edit. To finish a draft, send draftVideoId (the draft's output assetId) on seedance-2.5, with resolution 1080 (the default) or 2160 (4K) from the draft's draft.finishResolutions, to render the same shot, composition and motion at that native resolution. Prompt, references, length, aspect ratio and audio come from the draft, and sending them with draftVideoId is rejected. Only drafts that report draft.finishableUntil can be finished, within 7 days. Billed as a normal run of the draft's length and references at that resolution; a resolution the draft does not list is refused before any charge.
taskType"auto" | "reference" | "edit" | "extend"noOnly for seedance-2.5 and seedance-2.5-edit. Sent to BytePlus as omni_reference_task_type. Defaults to "auto", where Seedance decides from the prompt and references. "reference" generates a new video from the references; "edit" changes the reference video and needs aspectRatio and durationMode "auto"; "extend" continues it and needs aspectRatio "auto". Both need a referenceVideos item. Seedance still reads the prompt, and a prompt that reads as a different task type fails with TaskTypeMismatch.
seedintegernoOnly for seedance-2.5 and seedance-2.5-edit. Sent to BytePlus as seed, 0 to 2147483647. The same seed and inputs give a similar result, but Seedance does not guarantee an identical video.
retryAsEditbooleannoOnly for seedance-2.5. Defaults to false. With a referenceVideos item, Seedance can read the prompt as an edit, and an edit keeps the reference video's length and shape. When that differs from videoLength or aspectRatio, the process fails and says so. Send true to retry it once as an edit with durationMode and aspectRatio "auto" instead. The output then matches the reference video, and the retry is charged for the reference video's length before it runs; a team that can't cover that fails as out of credits. Any difference from the delivered length is settled when it finishes.
videoLengthnumbernoDuration in seconds.
startingFrameImageUrlstringnoImage URL used as the first frame. External URLs are auto-imported as assets first.
endingFrameImageUrlstringnoImage URL used as the last frame. External URLs are auto-imported as assets first.
durationMode"fixed" | "auto"noUse "auto" to match the duration of a supported reference video; defaults to "fixed".
generateAudiobooleannoWhether Seedance should generate a native audio track. Defaults to true. Must remain true when referenceAudios are attached.
bitrateMode"standard" | "high"noSeedance 2.x output bitrate preset. Use "high" for a crisper, larger file at the same credit cost; defaults to "standard".
keyframes{timeSeconds, imageUrl}[]noTimestamped image frames. timeSeconds is measured from the beginning of the generated clip.
referenceImages({assetId} | {url})[]noImage references. Send either {assetId} for a team asset or {url} for a public URL to import. Legacy {docId, url} objects remain accepted.
referenceVideos({assetId} | {url})[]noVideo references. Send either {assetId} for a team asset or {url} for a public URL to import. Legacy {docId, url} objects remain accepted.
referenceAudios({assetId} | {url})[]noAudio references. Send either {assetId} for a team asset or {url} for a public URL to import. Legacy {docId, url} objects remain accepted; private storage metadata is resolved server-side.
disableAutoRetriesbooleannoDisable automatic same-model retries. Defaults to false, so automatic retries are enabled.
customSettingsobjectnoThe model's own settings, the same ones the app's settings menu shows (for example GPT Image 2.6 quality, complexity and background: transparent). Only the settings listed in the model's capabilities.customSettings are accepted. Omitted settings keep their default.
endingFrameImageRefobjectno—
outputNamestringnoName for the output asset, up to 80 characters. It is stored as a lowercase hyphenated slug ("ORD-17 Concept A" becomes ord-17-concept-a), so use metadata for exact identifiers.
startingFrameImageRefobjectno—
stripAndReattachAudiobooleanno—

Referencing media: referenceImageUrls takes plain https image URLs. The normalized referenceImages, referenceVideos, and referenceAudios fields accept either {assetId} for an existing team asset or {url} for public media, which is imported automatically. Provider-native asset:// URIs are internal and are not accepted here. Legacy {docId, url} objects remain accepted for backwards compatibility.

Example

Add "dryRun": true to get the exact credit cost without starting, and always send an Idempotency-Key header on real starts.

POST /api/v1/processes
{
  "type": "video-generation",
  "model": "grok-imagine-v1.5",
  "input": {
    "prompt": "A slow dolly shot over a foggy pine forest at sunrise",
    "videoLength": 1,
    "aspectRatio": "16:9"
  }
}

Models at a glance

Credits are per generated output; ranges span the supported resolutions/lengths (exact prices in each model's section).

ModelCompanyCreditsLengths (s)Max ref imagesFramesAudio
genjutsuhiggsfielddryRun4–308—no
genjutsu-lighthiggsfielddryRun4–158—no
wan-3.0-primealibaba80–11702–3010start + endyes
wan-3.0alibaba60–7802–3010start + endyes
flux-3black-forest-labs310–12405–200start + endyes
minimax-h3minimax50–1605–159start + endyes
minimax-h3-maxminimax141–4245–159start + endyes
minimax-h3-max-turbominimax71–2125–15—start + endyes
minimax-h3-singularityminimax36–1045–159start + endyes
heygen-video-1heygen125–7885–309startyes
vidu-q4shengshu54–2883–1612startyes
seedance-2.5bytedance400–29204–3030start + endyes
seedance-2.5-flashbytedance60–3804–3010start + endyes
seedance-2.0bytedance220–8404–159start + endyes
grok-imaginexai58–1354, 6, 8, 10, 12, 14, 15—startyes
grok-imagine-v1.5xai20–2601–15—startyes
happy-horsealibaba90–4503–155startyes
seedance-2.0-fastbytedance30–1004–159start + endyes
seedance-2.0-minibytedance20–504–159start + endyes
wan-2.7alibaba30–1502, 4, 6, 8, 105start + endyes
pixverse-v6pixverse20–3701, 2, 4, 6, 8, 10, 12, 14, 15—startyes
kling-4.0kuaishou100–10103, 5, 8, 10, 12, 15, 20, 25, 3015start + endyes
kling-4.0-flashkuaishou60–4003, 5, 8, 10, 12, 15, 2015startyes
kling-3.0kuaishou72–2704, 6, 8, 10, 12, 14, 153start + endyes
kling-3.0-prokuaishou96–3604, 6, 8, 10, 12, 14, 153start + endyes
gemini-omni-flashgoogle115–3813–1010start + endyes
veo3-qualitygoogle21683start + endyes
veo3-fastgoogle8083start + endyes
seedance-2.0-openbytedance310–11804–159start + endyes
seedance-2.5-editbytedancedryRun—30—yes

Genjutsu (genjutsu)

Recast a source video's motion or swap a specific element using reference images.

CreditsCost depends on the source video duration and output resolution. POST /processes with dryRun: true for an exact quote.
Aspect ratiosauto — referenceVideos require "auto" (or omission)
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 720, 1080
Reference imagesup to 8
Elementsup to 8 (input.elements, by id or @ID)
Reference videosup to 1

Genjutsu Light (genjutsu-light)

A lower-cost Genjutsu for source clips up to 15 seconds: recast the motion or swap a specific element using reference images.

CreditsCost depends on the source video duration and output resolution. POST /processes with dryRun: true for an exact quote.
Aspect ratiosauto — referenceVideos require "auto" (or omission)
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720, 1080
Reference imagesup to 8
Elementsup to 8 (input.elements, by id or @ID)
Reference videosup to 1

WAN 3.0 Prime (wan-3.0-prime)

Alibaba's high-speed WAN 3.0 tier, around 7x faster with the same multimodal inputs and native audio.

Credits2s: 80 · 3s: 120 · 4s: 160 · 5s: 200 · 6s: 240 · 7s: 280 · 8s: 320 · 9s: 360 · 10s: 390 · 11s: 430 · 12s: 470 · 13s: 510 · 14s: 550 · 15s: 590 · 16s: 630 · 17s: 670 · 18s: 710 · 19s: 750 · 20s: 780 · 21s: 820 · 22s: 860 · 23s: 900 · 24s: 940 · 25s: 980 · 26s: 1020 · 27s: 1060 · 28s: 1100 · 29s: 1140 · 30s: 1170 (at 1080p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 720, 1080
customSettingsgenerate_audio_switch: true or false (default true)
enable_thinking: true or false (default false)
Reference imagesup to 10
Elementsup to 10 (input.elements, by id or @ID)
Reference videosup to 5 (15s total)
Reference audioup to 5 (15s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length5000

WAN 3.0 (wan-3.0)

Alibaba's newest multimodal video model with native audio, start/end frames, and image, video, and audio references.

Credits2s: 60 · 3s: 80 · 4s: 110 · 5s: 130 · 6s: 160 · 7s: 180 · 8s: 210 · 9s: 240 · 10s: 260 · 11s: 290 · 12s: 310 · 13s: 340 · 14s: 360 · 15s: 390 · 16s: 420 · 17s: 440 · 18s: 470 · 19s: 490 · 20s: 520 · 21s: 540 · 22s: 570 · 23s: 600 · 24s: 620 · 25s: 650 · 26s: 670 · 27s: 700 · 28s: 720 · 29s: 750 · 30s: 780 (at 1080p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 720, 1080
customSettingsgenerate_audio_switch: true or false (default true)
enable_thinking: true or false (default false)
Reference imagesup to 10
Elementsup to 10 (input.elements, by id or @ID)
Reference videosup to 5 (15s total)
Reference audioup to 5 (15s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length5000

FLUX.3 (flux-3)

Black Forest Labs' frontier video model for text, timestamped image keyframes, start/end-frame interpolation, or video continuation, with native audio and clips up to 20 seconds.

Credits5s: 310 · 6s: 370 · 7s: 440 · 8s: 500 · 9s: 560 · 10s: 620 · 11s: 680 · 12s: 750 · 13s: 810 · 14s: 870 · 15s: 930 · 16s: 990 · 17s: 1060 · 18s: 1120 · 19s: 1180 · 20s: 1240 (at 1080p)
Aspect ratios16:9, 9:16, 1:1, 21:9, 2:1, 4:3, 3:4, auto (default auto)
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20
Resolutions720, 1080, 1440, 2160
customSettingsgenerate_audio_switch: true or false (default true)
Reference imagesup to 0
Reference videosup to 1
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

MiniMax H3 (minimax-h3)

MiniMax's frontier 2K video model for text generation and multimodal image, video, and audio references.

Credits5s: 50 · 6s: 70 · 7s: 80 · 8s: 90 · 9s: 100 · 10s: 110 · 11s: 120 · 12s: 130 · 13s: 140 · 14s: 150 · 15s: 160 (at 768p)
Aspect ratios16:9, 21:9, 4:3, 1:1, 3:4, 9:16, adaptive
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions768, 1440
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length50000

MiniMax H3 Max (minimax-h3-max)

Fal's post-trained H3 variant for exceptionally fast 480p, 768p and 1080p video from text, first/end-frame animation, or image, video, and audio references, with native audio.

Credits5s: 141 · 6s: 169 · 7s: 198 · 8s: 226 · 9s: 254 · 10s: 282 · 11s: 311 · 12s: 339 · 13s: 367 · 14s: 395 · 15s: 424 (at 1080p)
Aspect ratios16:9, 21:9, 4:3, 1:1, 3:4, 9:16 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 768, 1080
customSettingsprompt_expansion_mode: disabled, balanced, quality (default balanced)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

MiniMax H3 Max Turbo (minimax-h3-max-turbo)

The fastest and cheapest H3 variant, generating 480p, 768p and 1080p text-to-video or first/end-frame animation with native audio in roughly half the time of H3 Max.

Credits5s: 71 · 6s: 85 · 7s: 99 · 8s: 113 · 9s: 127 · 10s: 141 · 11s: 155 · 12s: 169 · 13s: 184 · 14s: 198 · 15s: 212 (at 1080p)
Aspect ratios16:9, 21:9, 4:3, 1:1, 3:4, 9:16 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 768, 1080
customSettingsprompt_expansion_mode: disabled, balanced, quality (default balanced)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

MiniMax H3 Singularity (minimax-h3-singularity)

Community-tuned H3 with native audio. Choose Economy for a lower price or Fast for faster processing, with the same output settings.

Credits5s: 36 · 6s: 46 · 7s: 51 · 8s: 55 · 9s: 65 · 10s: 70 · 11s: 80 · 12s: 85 · 13s: 94 · 14s: 99 · 15s: 104 (at 1080p)
Aspect ratios16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720, 832, 1080, 1440, 2160
customSettingssingularity_speed: economy, fast (default fast)
singularity_seed: -1 to 281474976710655 (default -1)
singularity_first_steps: 1 to 64 (default 8)
singularity_refine_steps: 0 to 64 (default 4)
singularity_upscale: 1 to 4 (default 1.5)
singularity_base_megapixels: 0 to 16 (default 0)
singularity_denoise: 0.01 to 1 (default 1)
singularity_lora_strength: 0 to 2 (default 1)
singularity_sampler: euler, heun, dpmpp_2m, dpmpp_2m_sde (default euler)
singularity_scheduler: beta, normal, simple, karras, exponential, ddim_uniform (default beta)
singularity_reference_size: max, match (default max)
singularity_output_audio: true or false (default true)
singularity_reference_video_audio: true or false (default false)
singularity_crf: 0 to 51 (default 19)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (45s total)
Reference audioup to 3 (45s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

HeyGen Video (heygen-video-1)

HeyGen's post-trained H3 model for fast video up to 2K and 30 seconds from text, a first frame, or image, video, and audio references, with native audio.

Credits5s: 125 · 6s: 150 · 7s: 175 · 8s: 200 · 9s: 225 · 10s: 250 · 11s: 275 · 12s: 300 · 13s: 325 · 14s: 350 · 15s: 375 · 16s: 420 · 17s: 446 · 18s: 473 · 19s: 499 · 20s: 525 · 21s: 551 · 22s: 578 · 23s: 604 · 24s: 630 · 25s: 656 · 26s: 683 · 27s: 709 · 28s: 735 · 29s: 761 · 30s: 788 (at 1080p)
Aspect ratios16:9, 21:9, 4:3, 1:1, 3:4, 9:16 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 768, 1080, 1440
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationsupported
Start/end framesstart
Audio outputyes

Vidu Q4 (vidu-q4)

Vidu's reference-driven video model: keep up to 12 image references and 3 voices consistent, or start from text or a first frame. Native audio, 3 to 16 seconds, up to 4K.

Credits3s: 54 · 4s: 72 · 5s: 90 · 6s: 108 · 7s: 126 · 8s: 144 · 9s: 162 · 10s: 180 · 11s: 198 · 12s: 216 · 13s: 234 · 14s: 252 · 15s: 270 · 16s: 288 (at 1080p)
PricingLimited-time launch price; the future standard credit price is 1.43× the current rate
Aspect ratios16:9, 9:16, 4:3, 3:4, 1:1 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
Resolutions540, 720, 1080, 1440, 2160
Reference imagesup to 12
Elementsup to 12 (input.elements, by id or @ID)
Reference audioup to 3
Text-only generationsupported
Start/end framesstart
Audio outputyes
Max prompt length5000

Seedance 2.5 (seedance-2.5)

ByteDance's latest multimodal video model with native audio, 30-second generations, and expanded image, video, and audio references.

Credits4s: 400 · 5s: 490 · 6s: 590 · 7s: 690 · 8s: 780 · 9s: 880 · 10s: 980 · 11s: 1080 · 12s: 1170 · 13s: 1270 · 14s: 1370 · 15s: 1460 · 16s: 1560 · 17s: 1660 · 18s: 1750 · 19s: 1850 · 20s: 1950 · 21s: 2040 · 22s: 2140 · 23s: 2240 · 24s: 2340 · 25s: 2430 · 26s: 2530 · 27s: 2630 · 28s: 2720 · 29s: 2820 · 30s: 2920 (at 1080p)
PricingLimited-time launch price; the future standard credit price is 1.5× the current rate
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 720, 1080, 2160
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Draft mode480p runs are drafts unless draft: false; finish the same shot with draftVideoId on seedance-2.5 within 7 days, at resolution 1080 or 2160 (default 1080; each draft lists its own in draft.finishResolutions, and 2160 is 4K)
Reference imagesup to 30
Elementsup to 30 (input.elements, by id or @ID)
Reference videosup to 10 (30s total)
Reference audioup to 10 (30s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

Seedance 2.5 Flash (seedance-2.5-flash)

Faster, lower-cost Seedance 2.5 with native audio, clips up to 30 seconds, and image, video, and audio references at 480p or 720p.

Credits4s: 60 · 5s: 70 · 6s: 80 · 7s: 90 · 8s: 110 · 9s: 120 · 10s: 130 · 11s: 140 · 12s: 160 · 13s: 170 · 14s: 180 · 15s: 190 · 16s: 200 · 17s: 220 · 18s: 230 · 19s: 240 · 20s: 250 · 21s: 270 · 22s: 280 · 23s: 290 · 24s: 300 · 25s: 320 · 26s: 330 · 27s: 340 · 28s: 350 · 29s: 370 · 30s: 380 (at 480p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
Resolutions480, 720
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Reference imagesup to 10
Elementsup to 10 (input.elements, by id or @ID)
Reference videosup to 5 (30s total)
Reference audioup to 5 (30s total)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

Seedance 2.0 (seedance-2.0)

ByteDance's video model with multi-modal input, lip sync, and multi-shot narrative.

Credits4s: 220 · 5s: 280 · 6s: 340 · 7s: 390 · 8s: 450 · 9s: 510 · 10s: 560 · 11s: 620 · 12s: 670 · 13s: 730 · 14s: 790 · 15s: 840 (at 1080p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720, 1080, 2160
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationnot supported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

Grok Imagine (grok-imagine)

Grok's best and latest video model with sound included.

Credits4s: 58 · 6s: 72 · 8s: 86 · 10s: 100 · 12s: 114 · 14s: 128 · 15s: 135 (at 720p)
Aspect ratios16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:1 — "auto" also accepted with startingFrameImageUrl
Lengths (s)4, 6, 8, 10, 12, 14, 15
Resolutions720
Text-only generationsupported
Start/end framesstart
Audio outputyes
Max prompt length4096

Grok Imagine 1.5 (grok-imagine-v1.5)

xAI's Grok Imagine 1.5 image-to-video model with audio included.

Credits1s: 20 · 2s: 40 · 3s: 50 · 4s: 70 · 5s: 90 · 6s: 110 · 7s: 120 · 8s: 140 · 9s: 160 · 10s: 170 · 11s: 190 · 12s: 210 · 13s: 230 · 14s: 240 · 15s: 260 (at 480p)
Aspect ratios16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:1 — "auto" also accepted with startingFrameImageUrl
Lengths (s)1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720
Text-only generationnot supported
Start/end framesstart
Audio outputyes
Max prompt length4096

Happy Horse (happy-horse)

Alibaba's reference-to-video model with native audio and 720p output.

Credits3s: 90 · 4s: 120 · 5s: 150 · 6s: 180 · 7s: 210 · 8s: 240 · 9s: 270 · 10s: 300 · 11s: 330 · 12s: 360 · 13s: 390 · 14s: 420 · 15s: 450 (at 720p)
Aspect ratios16:9, 9:16, 1:1, 4:3, 3:4
Lengths (s)3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions720
Reference imagesup to 5
Elementsup to 5 (input.elements, by id or @ID)
Text-only generationnot supported
Start/end framesstart
Audio outputyes
Max prompt length5000

Seedance 2.0 Fast (seedance-2.0-fast)

Fast variant of ByteDance's video model with multi-modal input, lip sync, and multi-shot narrative.

Credits4s: 30 · 5s: 40 · 6s: 40 · 7s: 50 · 8s: 50 · 9s: 60 · 10s: 70 · 11s: 70 · 12s: 80 · 13s: 80 · 14s: 90 · 15s: 100 (at 480p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationnot supported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

Seedance 2.0 Mini (seedance-2.0-mini)

ByteDance's economical video model for fast multimodal drafts with native audio.

Credits4s: 20 · 5s: 20 · 6s: 20 · 7s: 30 · 8s: 30 · 9s: 30 · 10s: 40 · 11s: 40 · 12s: 40 · 13s: 40 · 14s: 50 · 15s: 50 (at 480p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationnot supported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

WAN 2.7 (wan-2.7)

Wan's latest reference-driven video model with image and video guidance for better consistency.

Credits2s: 30 · 4s: 60 · 6s: 90 · 8s: 120 · 10s: 150 (at 1080p)
Aspect ratios16:9, 9:16, 1:1, 4:3, 3:4
Lengths (s)2, 4, 6, 8, 10
Resolutions720, 1080
Reference imagesup to 5
Elementsup to 5 (input.elements, by id or @ID)
Reference videosup to 1
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length5000

PixVerse V6 (pixverse-v6)

PixVerse's latest Fal-hosted video model with optional audio and first-frame image support.

Credits1s: 20 · 2s: 50 · 4s: 100 · 6s: 150 · 8s: 200 · 10s: 250 · 12s: 300 · 14s: 350 · 15s: 370 (at 1080p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 2:3, 3:2, 21:9 — startingFrameImageUrl requires "auto" (or omission)
Lengths (s)1, 2, 4, 6, 8, 10, 12, 14, 15
Resolutions360, 540, 720, 1080
customSettingsgenerate_audio_switch: true or false (default true)
generate_multi_clip_switch: true or false (default true)
Text-only generationsupported
Start/end framesstart
Audio outputyes
Max prompt length2048

Kling 4.0 (kling-4.0)

Kling's newest video model: up to 30 seconds at 1080p with native audio, start/end frames, image and element references, and video editing.

Credits3s: 100 · 5s: 170 · 8s: 270 · 10s: 340 · 12s: 400 · 15s: 500 · 20s: 670 · 25s: 840 · 30s: 1010 (at 1080p)
Aspect ratios16:9, 9:16, 1:1, 21:9 — referenceVideos require "auto" (or omission)
Lengths (s)3, 5, 8, 10, 12, 15, 20, 25, 30
Resolutions720, 1080
Reference imagesup to 15
Elementsup to 7 (input.elements, by id or @ID)
Reference videosup to 1
Text-only generationsupported
Start/end framesstart + end
Audio outputyes
Max prompt length2500

Kling 4.0 Flash (kling-4.0-flash)

Faster, lower-cost Kling 4.0: up to 20 seconds at 720p with native audio, a start frame, image and element references, and video editing.

Credits3s: 60 · 5s: 100 · 8s: 160 · 10s: 200 · 12s: 240 · 15s: 300 · 20s: 400 (at 720p)
Aspect ratios16:9, 9:16, 1:1, 21:9 — startingFrameImageUrl and referenceVideos require "auto" (or omission)
Lengths (s)3, 5, 8, 10, 12, 15, 20
Resolutions720
Reference imagesup to 15
Elementsup to 7 (input.elements, by id or @ID)
Reference videosup to 1
Text-only generationsupported
Start/end framesstart
Audio outputyes
Max prompt length2500

Kling 3.0 (kling-3.0)

Video model with start/end frames, references, and native audio.

Credits4s: 72 · 6s: 108 · 8s: 144 · 10s: 180 · 12s: 216 · 14s: 252 · 15s: 270 (at 720p)
Aspect ratios16:9, 9:16, 1:1 — startingFrameImageUrl and referenceVideos require "auto" (or omission)
Lengths (s)4, 6, 8, 10, 12, 14, 15
Resolutions720
customSettingskeepAudio: true or false (default true)
Reference imagesup to 3
Elementsup to 3 (input.elements, by id or @ID)
Reference videosup to 1
Start/end framesstart + end
Audio outputyes
Max prompt length2500

Kling 3.0 Pro (kling-3.0-pro)

Higher-cost Kling model with start/end frame, reference, and native audio support.

Credits4s: 96 · 6s: 144 · 8s: 192 · 10s: 240 · 12s: 288 · 14s: 336 · 15s: 360 (at 1080p)
Aspect ratios16:9, 9:16, 1:1 — startingFrameImageUrl and referenceVideos require "auto" (or omission)
Lengths (s)4, 6, 8, 10, 12, 14, 15
Resolutions1080
customSettingskeepAudio: true or false (default true)
Reference imagesup to 3
Elementsup to 3 (input.elements, by id or @ID)
Reference videosup to 1
Start/end framesstart + end
Audio outputyes
Max prompt length2500

Gemini Omni 1.1 Flash (gemini-omni-flash)

Google's fast multimodal video model with native audio, first/end-frame interpolation, video editing and extension, and output up to 4K.

Credits3s: 115 · 4s: 153 · 5s: 191 · 6s: 229 · 7s: 267 · 8s: 305 · 9s: 343 · 10s: 381 (at 1080p)
Aspect ratios16:9, 9:16 — referenceVideos require "auto" (or omission)
Lengths (s)3, 4, 5, 6, 7, 8, 9, 10
Resolutions360, 720, 1080, 2160
Reference imagesup to 10
Reference videosup to 1
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

Veo3.1 Quality (veo3-quality)

Google's highest quality video model with sound included. Blocks celebrities.

Credits8s: 216 (at 1080p)
Aspect ratios16:9, 9:16
Lengths (s)8
Resolutions1080
Reference imagesup to 3
Elementsup to 3 (input.elements, by id or @ID)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

Veo3.1 (veo3-fast)

Google's latest video model with sound included. Blocks celebrities.

Credits8s: 80 (at 1080p)
Aspect ratios16:9, 9:16
Lengths (s)8
Resolutions1080
Reference imagesup to 3
Elementsup to 3 (input.elements, by id or @ID)
Text-only generationsupported
Start/end framesstart + end
Audio outputyes

Seedance 2.0 Open (seedance-2.0-open)

Less restricted Seedance 2.0 variant for content that the standard model may block.

Credits4s: 310 · 5s: 390 · 6s: 480 · 7s: 550 · 8s: 630 · 9s: 710 · 10s: 780 · 11s: 870 · 12s: 940 · 13s: 1020 · 14s: 1110 · 15s: 1180 (at 1080p)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, adaptive
Lengths (s)4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
Resolutions480, 720, 1080, 2160
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Reference imagesup to 9
Elementsup to 9 (input.elements, by id or @ID)
Reference videosup to 3 (15s total)
Reference audioup to 3 (15s total)
Text-only generationnot supported
Start/end framesstart + end
Audio outputyes
Max prompt length20000

Seedance 2.5 Edit (seedance-2.5-edit)

Dedicated Seedance 2.5 base-video workflow for editing or extending a video, or generating a natural sequel or prequel. The first referenceVideos item is the base video; duration and aspect ratio are automatic. Prompt is optional for extend, sequel, and prequel, but required for edit.

CreditsCost depends on the base video's automatic output duration — POST /processes with dryRun: true for an exact quote.
PricingLimited-time launch price; the future standard credit price is 1.5× the current rate
Base videorequired as the first referenceVideos item (array index 0)
Operationsedit, extend, sequel, prequel
Promptrequired for edit; optional for extend and sequel and prequel
Durationautomatic (durationMode: "auto")
Aspect ratiosauto (default auto) — referenceVideos require "auto" (or omission)
Resolutions480, 720, 1080, 2160
Bitrate modesstandard, high (default standard)
Generated audioconfigurable (default on; required with reference audio)
Draft mode480p runs are drafts unless draft: false; finish the same shot with draftVideoId on seedance-2.5 within 7 days, at resolution 1080 or 2160 (default 1080; each draft lists its own in draft.finishResolutions, and 2160 is 4K)
Reference imagesup to 30
Reference videosup to 10 (30s total)
Reference audioup to 10 (30s total)
Text-only generationnot supported
Audio outputyes
Max prompt length20000

On this page