Video Generation API

The Video Generation API lets you create high-quality videos from a text prompt (text-to-video) or from a reference image (image-to-video) through one unified endpoint:

POST /api/v1/videos/generations

The request is dispatched by model. Most models take an optional image_url for a first-frame image and run text-to-video when it is omitted. Multi-reference models (viduq3_turbo_reference, viduq4_preview_reference, grok_imagine_refs, grok_imagine_1_5_refs) take an image_urls array instead. Single-image endpoints — including grok_imagine and grok_imagine_1_5 — take a scalar image_url, and for both of them it is optional — the two fields are not interchangeable. See each model page for its exact parameters.

Available Models

Model	Best for	Resolution	Duration	Native Audio
`seedance-2-0`	Highest-quality Seedance output	480p / 720p / 1080p / 4k	4-15s	✅
`seedance-2-0-fast`	Faster drafts and iteration	480p / 720p	4-15s	✅
`seedance-2-0-mini`	Lowest-cost Seedance iteration	480p / 720p	4-15s	✅
`wan22_spicy`	Cost-effective text-to-video and image-to-video	480p / 720p	Model default	❌
`kling-3-0`	Highest quality, full feature set	720p / 1080p	3-15s	✅ `sound: on/off`
`kling-3-0-turbo`	Faster generation	720p / 1080p	3-15s	❌
`kling-2-5-turbo`	Cost-effective	720p / 1080p	5s or 10s	❌
`viduq3_pro`	Highest-quality Vidu output	540p / 720p / 1080p	1-16s	✅ (on by default)
`viduq3_turbo`	Faster, lower-cost Vidu output	540p / 720p / 1080p	1-16s	✅ (on by default)
`viduq3_turbo_reference`	Consistent subjects from reference images	540p / 720p / 1080p	3-16s	✅ (on by default)
`viduq4_preview`	Newest Vidu image-to-video, up to 4K	540p / 720p / 1080p / 2K / 4K	3-16s	✅ `generate_audio` (on by default)
`viduq4_preview_reference`	Up to 15 reference images plus reference voices, up to 4K	540p / 720p / 1080p / 2K / 4K	3-16s	✅ `generate_audio` (on by default)
`hailuo_2_3`	Sharp motion and camera control	768P / 1080P	6s or 10s	❌
`hailuo_2_3_fast`	Lower-cost Hailuo drafts	768P / 1080P	6s or 10s	❌
`grok_imagine`	Text-to-video and one reference image	480p / 720p / 1080p	6-30s	❌
`grok_imagine_refs`	Animating 1-7 reference images	480p / 720p	6-30s	❌
`grok_imagine_1_5`	Text-to-video and one reference image	480p / 720p / 1080p	1-15s	❌
`grok_imagine_1_5_refs`	Animating 1-7 reference images	480p / 720p / 1080p	1-15s	❌
`pix_v6`	PixVerse text/image/transition video	360p / 540p / 720p / 1080p	5 / 8 / 10 / 15s	✅ `sound: on/off`
`gemini_omni_video`	Image-to-video with synchronized audio	720p / 1080p / 4k	4 / 6 / 8 / 10s	✅ (always on)
`ai-talking-avatar`	Lip-syncing a portrait to an audio track	540p / 720p	Follows your audio (3-60s), or `duration` to cut it shorter	✅ (your audio)
`ai-talking-avatar-video`	Making the person in a video speak your audio, with head and expression motion	540p / 720p	Follows your audio and video (3-60s), or `duration` to cut it shorter	✅ (your audio)
`lipsync-video`	Re-animating the person in an existing video to speak your audio	Follows your video (540p / 720p / 1080p)	Follows your audio (420s max, 300s at 1080p)	✅ (your audio)
`video-upscaler`	Upscaling an existing clip to a higher resolution	720p / 1080p / 2k / 4k	Follows your video (60s max, 30s for 4k)	✅ (kept from source)
`anime-upscaler`	Upscaling an anime clip — tuned for line art and flat colour	1080p / 2k / 4k	Follows your video (180s max)	✅ (kept from source)
`video-watermark-remover`	Removing watermarks and logos from an existing clip, including moving ones	Follows your video (up to 1080p)	Follows your video (60s max)	✅ (kept from source)

Model values are not uniformly punctuated: Seedance and Kling use hyphens (seedance-2-0, kling-3-0), while Vidu, Hailuo, Grok, PixVerse and Wan use underscores (viduq3_pro, hailuo_2_3, grok_imagine, pix_v6, wan22_spicy). Copy the value exactly as shown — anything else returns Unsupported model.

Choosing a Vidu model

viduq3_pro and viduq3_turbo cover text-to-video, image-to-video and start-end frame generation — Turbo is roughly half the price, Pro has the higher fidelity. Reference-image generation is a separate model, viduq3_turbo_reference, because Vidu prices it differently and enforces a 3-second minimum. The two are mutually exclusive: viduq3_turbo rejects image_urls, and viduq3_turbo_reference rejects image_url / image_end_url. Vidu Q4 Preview is the newest generation, with output up to 4K. It comes as two models with the same price: viduq4_preview turns a first-frame image_url into a video (it has no text-to-video or start-end frame mode, so image_url is required), and viduq4_preview_reference takes 1-15 image_urls plus up to 3 optional ref_audios clips for voices and dialogue.

Feature Matrix — Wan & Kling

Feature support for the other families is listed on their own pages.

Feature	wan22_spicy	kling-3-0	kling-3-0-turbo	kling-2-5-turbo
Text-to-video	✅	✅	✅	✅
Image-to-video (first frame)	✅	✅	✅	✅
First & last frame (`image_end_url`)	❌	✅	❌	✅
Native audio (`sound`)	❌	✅	❌	❌
Negative prompt	❌	✅	❌	✅
Multi-shot (`model_params.multi_shot`)	❌	✅	❌	❌

Pricing

Every price below is given in tokens and USD, converted at $0.02/token (10 USD = 500 tokens). Tokens are what is actually deducted from your balance; the USD figure is the same price expressed in money. Seedance is charged per second of generated video:

Model	480p	720p	1080p	4k
`seedance-2-0`	5.98 /s ($0.120)	12.94 /s ($0.259)	32.24 /s ($0.645)	65.78 /s ($1.316)
`seedance-2-0-fast`	4.81 /s ($0.096)	10.47 /s ($0.209)	—	—
`seedance-2-0-mini`	2.99 /s ($0.060)	6.44 /s ($0.129)	—	—

Synchronized audio is included at no additional charge. content_filter does not affect pricing. Wan 2.2 has a fixed per-task cost:

Model	480p	720p
`wan22_spicy`	5 tokens/task ($0.10)	10 tokens/task ($0.20)

Kling costs are charged per second of generated video:

Model	720p	1080p
`kling-3-0` (sound off)	4.2 tokens/s ($0.084)	5.6 tokens/s ($0.112)
`kling-3-0` (sound on)	6.3 tokens/s ($0.126)	8.4 tokens/s ($0.168)
`kling-3-0-turbo`	5.6 tokens/s ($0.112)	7 tokens/s ($0.140)
`kling-2-5-turbo`	1.89 tokens/s ($0.038)	3.15 tokens/s ($0.063)

Example: a 5-second kling-2-5-turbo video at 720p costs 9.45 tokens (= $0.189). Vidu costs are charged per second of generated video:

Model	540p	720p	1080p
`viduq3_pro`	2.25 tokens/s ($0.045)	5 tokens/s ($0.10)	6 tokens/s ($0.12)
`viduq3_turbo`	1.75 tokens/s ($0.035)	2.75 tokens/s ($0.055)	3.25 tokens/s ($0.065)
`viduq3_turbo_reference`	1 token/s ($0.02)	2.5 tokens/s ($0.05)	3.25 tokens/s ($0.065)

Example: an 8-second viduq3_turbo video at 720p costs 22 tokens (= $0.44). Vidu Q4 Preview (viduq4_preview and viduq4_preview_reference, same price):

540p	720p	1080p	2K	4K
2.925 tokens/s ($0.0585)	6.175 tokens/s ($0.1235)	7.8 tokens/s ($0.156)	12.35 tokens/s ($0.247)	25.35 tokens/s ($0.507)

Example: a 5-second viduq4_preview video at 1080p costs 39 tokens (= $0.78). Hailuo has a fixed cost per video, by resolution and duration:

Model	768P 6s	768P 10s	1080P 6s
`hailuo_2_3`	18.2 tokens ($0.364)	36.4 tokens ($0.728)	31.85 tokens ($0.637)
`hailuo_2_3_fast`	12.35 tokens ($0.247)	20.8 tokens ($0.416)	21.45 tokens ($0.429)

1080P is only available at 6 seconds. Grok Imagine costs are charged per second of generated video. All four endpoints share the same 480p and 720p rates; at 1080p, both Grok Imagine 1.5 endpoints cost twice their 720p rate:

Model	480p	720p	1080p
`grok_imagine`	2.5 tokens/s ($0.05)	4.7 tokens/s ($0.094)	8.3 tokens/s ($0.166)
`grok_imagine_refs`	2.5 tokens/s ($0.05)	4.7 tokens/s ($0.094)	not available
`grok_imagine_1_5`	2.5 tokens/s ($0.05)	4.7 tokens/s ($0.094)	9.4 tokens/s ($0.188)
`grok_imagine_1_5_refs`	2.5 tokens/s ($0.05)	4.7 tokens/s ($0.094)	9.4 tokens/s ($0.188)

Example: a 6-second grok_imagine video at 720p costs 28.2 tokens (=

0.564). An 8-second `grok_imagine_1_5` video at 720p costs 37.6 tokens (=

0.752). An 8-second grok_imagine_1_5 video at 1080p costs 75.2 tokens (= $1.504). Grok Imagine offers 1080p only on the single-image endpoint grok_imagine; grok_imagine_refs runs at 480p or 720p. Both Grok Imagine 1.5 endpoints offer 1080p with any supported number of reference images. 1080p videos take longer to finish than 480p or 720p. PixVerse costs are charged per second of generated video. Native audio (sound: on) uses the higher rate:

Model	360p	540p	720p	1080p
`pix_v6` (silent)	1.57 tokens/s ($0.0314)	2.19 tokens/s ($0.0438)	2.82 tokens/s ($0.0564)	5.64 tokens/s ($0.1128)
`pix_v6` (audio)	2.19 tokens/s ($0.0438)	2.82 tokens/s ($0.0564)	3.76 tokens/s ($0.0752)	7.21 tokens/s ($0.1442)

Example: a 5-second pix_v6 video at 720p costs 14.11 tokens silent (=

0.2822). The same video with audio costs 18.81 tokens (=

0.3762). Gemini Omni Video is charged per generation by resolution and duration (720p and 1080p share the same price):

resolution	4s	6s	8s	10s
`gemini_omni_video` 720p / 1080p	20.48 ($0.4096)	27.3 ($0.546)	34.13 ($0.6826)	40.95 ($0.819)
`gemini_omni_video` 4k	47.78 ($0.9556)	54.6 ($1.092)	61.43 ($1.2286)	68.25 ($1.365)

Check your balance with Get User Tokens.

How It Works

Submit a task — call POST /api/v1/videos/generations with your model and prompt
Get the task id — the response returns an img_uuid
Get the result — poll Get Job Detail with the img_uuid, or pass a custom_callback_url to receive status notifications automatically

Example

curl -X POST 'https://api.goenhance.ai/api/v1/videos/generations' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "kling-3-0",
    "prompt": "The cat slowly turns its head and blinks",
    "image_url": "https://your-cdn.com/cat.jpg",
    "duration": 5,
    "resolution": "1080p",
    "sound": "on"
  }'

Response:

{
  "code": 0,
  "msg": "Success",
  "data": {
    "img_uuid": "c12b656c-747a-44fd-9c80-add79b0c52d5"
  }
}

See the per-model pages for the full parameter reference: Seedance 2.0, Seedance 2.0 Fast, Seedance 2.0 Mini, Wan 2.2, Kling 3.0, Kling 3.0 Turbo, Kling 2.5 Turbo, Vidu Q3 Pro, Vidu Q3 Turbo, Vidu Q3 Turbo Reference, Vidu Q4 Preview, Vidu Q4 Preview Reference, Hailuo 2.3, Hailuo 2.3 Fast, Grok Imagine, Grok Imagine 1.5, PixVerse V6, Gemini Omni Video.

Getting started

API

Video Generation API Overview

Video Generation API

Available Models

Choosing a Vidu model

Feature Matrix — Wan & Kling

Pricing

How It Works

Example

​Video Generation API

​Available Models

​Choosing a Vidu model

​Feature Matrix — Wan & Kling

​Pricing

​How It Works

​Example

Video Generation API

Available Models

Choosing a Vidu model

Feature Matrix — Wan & Kling

Pricing

How It Works

Example