Models API

List all video and image generation models, inspect their capabilities and supported parameters, and run one-off playground generations. Use model details to configure jobs correctly via the Jobs API.

Model Families

U-Gen supports 5 model families across video and image generation. Each model has unique capabilities, aspect ratios, and billing.

Veo 3.1

Google DeepMind

CategoryVideo

Aspects9:16, 16:9, Auto

Duration8s fixed

TiersFast (standard) / Quality (premium)

veo3_fastveo3

UGC Studio + Playground

Kling 3.0

Kuaishou

CategoryVideo

Aspects9:16, 16:9, 1:1

Duration3–10s flexible

TiersStandard / Pro (premium)

kling3_stdkling3_pro

UGC Studio + Playground

Sora 2

OpenAI

CategoryVideo

Aspects9:16, 16:9

Duration10s or 15s

TiersStandard

sora-2-text-to-videosora-2-image-to-video

Playground only

Sora 2 Pro

OpenAI

CategoryVideo

Aspects9:16, 16:9

Duration10s, 15s, or 25s (storyboard)

TiersPremium

sora-2-pro-text-to-videosora-2-pro-image-to-videosora2_storyboard

Playground only

NanoBanana 2

Google DeepMind

CategoryImage

Aspects15 ratios + Auto

DurationN/A (images)

TiersStandard

nano_banana_2

Keyframe generation + Playground

UGC Studio vs Playground

The Jobs API (UGC Studio) only accepts veo3_fast, veo3, kling3_std, and kling3_pro. Sora 2, Sora 2 Pro, and NanoBanana models are available through the Playground API (POST /models/{id}/try) for one-off generations.

List Models

Retrieve all video and image generation models with their capabilities. Returns models that are currently available or in beta.

GET/api/v1/models

• Returns a bare array (not paginated). Currently 11 models across 5 families.

• Models with status "coming_soon" are excluded from the response.

• The capabilities object describes what each model supports: aspect ratios, durations, keyframes, sound, and resolution upgrades.

• The default_variant field maps to the pricing table for credit estimation.

Request Example

curl -X GET https://api.u-gen.ai/api/v1/models \
  -H "X-API-Key: YOUR_API_KEY"

Get Model Details

Retrieve full details for a single model by its ID.

GET/api/v1/models/{model_id}

• Returns 404 if the model ID is not found in the registry.

• Same schema as a single entry in the List Models response.

Path

Parameter	Type	Description
`model_id`REQ	string	The model identifier.

Request Example

curl -X GET https://api.u-gen.ai/api/v1/models/veo3_fast \
  -H "X-API-Key: YOUR_API_KEY"

Capability Fields

Use capabilities.supported_aspect_ratios and capabilities.supported_durations to populate your UI dropdowns dynamically. Kling models support supports_sound and supports_elements for built-in audio and character references.

Try Model (Playground)

Start a one-off generation using any available model. Returns a task ID immediately — poll the task status endpoint for results.

POST/api/v1/models/{model_id}/try

• Returns immediately with a task_id. Poll GET /models/tasks/{task_id} for status.

• Task state is stored in memory and lost on server restart (playground use only).

• Aspect ratio is validated against the model’s capabilities.supported_aspect_ratios.

• Reference-to-video is only available with veo3_fast and does not support the Auto aspect ratio.

• Kling prompt limit is 2,500 characters. Sora prompt limit is 10,000 characters.

• For upscale operations, use model_id "veo_get_1080p" or "veo_get_4k" with the original task_id.

Path

Parameter	Type	Description
`model_id`REQ	string	The model to generate with.

Body

Parameter	Type	Description
`model_id`REQ	string	Must match the path parameter.
`generation_type`	string	Generation mode. text_to_videoimage_to_videoreference_to_video
`prompt`	string	Text prompt describing the desired output. Max 10,000 chars (Sora), 2,500 chars (Kling)
`aspect_ratio`	string Default: 9:16	Output aspect ratio. Must be supported by the model.
`start_image_url`	string	Start frame image URL for image-to-video.
`end_image_url`	string	Optional end frame image URL.
`reference_images`	string[]	Reference image URLs for reference-to-video (Veo 3.1 Fast only, 1–3 images).
`duration`	integer	Duration in seconds. Sora: 10, 15, or 25. Kling: 3–15.

Body (Sora)

Parameter	Type	Description
`shots`	ShotItem[]	Scene array for Sora Storyboard. Each item: { prompt, duration }.
`size`	string	Resolution toggle for Sora Pro. standardhigh

Body (Kling)

Parameter	Type	Description
`kling_mode`	string	Override Kling quality tier. stdpro
`sound`	boolean	Enable Kling built-in sound effects.
`kling_elements`	object[]	Element references for character consistency. Each: { name, description?, element_input_urls? }.

Body (Image Gen)

Parameter	Type	Description
`resolution`	string	Image resolution for NanoBanana. 1K2K4K
`output_format`	string	Image output format for NanoBanana. pngjpg

Body (Upscale)

Parameter	Type	Description
`task_id`	string	Required for upscale operations (veo_get_1080p / veo_get_4k model IDs).

Request Example

curl -X POST https://api.u-gen.ai/api/v1/models/veo3_fast/try \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_id": "veo3_fast",
    "generation_type": "image_to_video",
    "start_image_url": "https://example.com/product.jpg",
    "prompt": "A woman holds the product in a sunlit studio, examining it closely",
    "aspect_ratio": "9:16"
  }'

Poll Task Status

Check the status of a playground generation or upscale task. Poll until status is completed or failed.

GET/api/v1/models/tasks/{task_id}

• Status values: queued, processing, completed, failed.

• output_url is populated only when status is "completed".

• error_message is populated only when status is "failed".

• Recommended polling interval: every 3–5 seconds.

Path

Parameter	Type	Description
`task_id`REQ	string	The task ID returned by the Try Model endpoint.

Request Example

curl -X GET https://api.u-gen.ai/api/v1/models/tasks/a1b2c3d4-e5f6-7890-abcd-ef1234567890 \
  -H "X-API-Key: YOUR_API_KEY"

Reference-to-Video Restrictions

Veo 3.1 reference_to_video is only available on the veo3_fast model. The Auto aspect ratio is not supported in this mode. Provide 1–3 reference images via the reference_images field.

Kling Language Support

Kling 3.0 models support 5 languages: Chinese (zh), English (en), Japanese (ja), Korean (ko), and Spanish (es). In the UGC Studio, selecting an unsupported language automatically switches the model to Veo 3.1 Fast.