Seedance 2 ProReve 2.0

Image Models

Generate and edit images with Reve 2.0, Nano Banana, Nano Banana Pro and Seedream 5.

Image models run through POST /api/v1/generate. Providing image_urls switches a request to image-to-image (edit) mode.

Models

modelModesBase credits / image
reve-2.0text-to-image, image-to-image12 (text) / 20 (edit)
nano-bananatext-to-image20
nano-banana-protext-to-image, image-to-image70 (×2 at 4K)
seedream-5text-to-image, image-to-image25

reve-2.0 is the default. Credits are multiplied by num_images; Reve test_time_scaling and paid postprocessing operations increase its cost.

Parameters

FieldTypeRequiredDescription
modelstringNoDefault reve-2.0.
promptstringYesText description of the image.
image_urlsstring[]NoReference image(s) → switches to image-to-image / edit mode.
aspect_ratiostringNoe.g. 1:1, 16:9, 3:2, 4:3, 3:4, 2:3, 9:16, or auto for Reve.
num_imagesintegerNoNumber of images (models that support it). Default 1.
output_formatstringNopng, jpeg, or webp.
resolutionstringNo1K, 2K, 4K (model-dependent). Reve currently uses detail scaling and post-generation sizing rather than this parameter.
test_time_scalingintegerNoReve only. 1–15. Higher = more detail; billed as base × N.
versionstringNoReve v1 only. latest, latest-fast, or dated Reve v1 versions where supported.
upscalestringNoReve only. enabled or disabled. Enables post-generation upscale.
upscale_factorintegerNoReve only. 2, 3, or 4 when upscale is enabled.
remove_backgroundbooleanNoReve only. Applies background removal after generation.
postprocessingobject[]NoReve only. Supports upscale, remove_background, fit_image, and effect; current UI exposes upscale/background.

Reve 2.0 Capability Status

Screenshot featureCurrent statusNotes
Generate and edit images in 4K nativelyPartially supportedReve create/edit/remix are integrated with test_time_scaling, 2x/3x/4x upscale, and output format controls. Reve also documents fit_image postprocessing with max_dim, max_width, or max_height up to 4096, but this is postprocessing, not a separate native 4K resolution switch in our current API.
Photorealism and textSupported as model capabilityReve 2.0 is used for text-to-image and image-to-image. Users steer photorealism and text rendering through the prompt; there is no separate photorealism or typography-strength parameter exposed by Reve.
Structured layoutsOfficially supported by Reve v2, not yet integrated hereReve v2 create accepts description; Reve v2 edit accepts old_description and new_description. The layout object contains prompt plus regions with label, prompt, normalized bbox, optional preserve, image_index, and image_region_index. Our current provider still calls v1 create/edit/remix.
Use references for strong style adherenceSupported, with a v2 upgrade pathCurrent image-to-image supports one or more reference images: single image routes to Reve v1 edit and multiple images route to Reve v1 remix. Reve v2 also supports structured references entries with an image object and optional reference prompt; this is not yet integrated.

Reve v2 Layouts

Reve's structured layout API is available on the v2 endpoints, not as a structured_layout field on the current v1 requests.

EndpointLayout fields
POST /v2/image/createinstruction, optional references, optional description, aspect_ratio, postprocessing
POST /v2/image/editinstruction, image, optional references, optional old_description, optional new_description, aspect_ratio, postprocessing

Layout descriptions use normalized coordinates:

{
  "prompt": "A product poster with a headline and a centered bottle",
  "regions": [
    {
      "label": "headline",
      "prompt": "Large readable headline text at the top",
      "bbox": { "x0": 0.08, "y0": 0.06, "x1": 0.92, "y1": 0.22 }
    },
    {
      "label": "bottle",
      "prompt": "A glossy glass bottle centered in the composition",
      "bbox": { "x0": 0.34, "y0": 0.28, "x1": 0.66, "y1": 0.82 }
    }
  ]
}

Text-to-image

curl -X POST https://reve2img.com/api/v1/generate \
  -H "Authorization: Bearer sk-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "reve-2.0",
    "prompt": "a red panda astronaut, cinematic lighting",
    "aspect_ratio": "16:9",
    "test_time_scaling": 2,
    "upscale": "enabled",
    "upscale_factor": 2
  }'

Image-to-image (edit)

curl -X POST https://reve2img.com/api/v1/generate \
  -H "Authorization: Bearer sk-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-pro",
    "prompt": "make the sky a dramatic sunset",
    "image_urls": ["https://example.com/photo.jpg"]
  }'

Then poll /api/v1/status until SUCCESS.

API Reference

Base URL, authentication, endpoints, response format and error codes for the API.

Video Models

Generate video with HappyHorse 1.0, Kling 3.0, Omni Video, Grok Imagine and Kling 2.6 Motion Control.

On this page

Models
Parameters
Reve 2.0 Capability Status
Reve v2 Layouts
Text-to-image
Image-to-image (edit)