Grok Imagine Image (Speed)

Generate and edit images using the speed-focused Grok Imagine image model. The API produces exactly one output image per task.

The enable_pro parameter has been removed. Speed and quality are now separate models: use xai/grok-imagine-image for speed or xai/grok-imagine-2.0-image for Grok Imagine 2.0 quality.

Model

xai/grok-imagine-image

Parameters

Parameter	Type	Required	Default	Description
`prompt`	string	Yes	—	Text description or edit instruction
`aspect_ratio`	string	No	`"1:1"`	`"1:1"`, `"2:3"`, `"3:2"`, `"9:16"`, or `"16:9"`
`image_urls`	string[]	No	`null`	1–5 reference image URLs. Providing this field activates edit mode
`upsample_prompt`	boolean	No	`true`	Let AI enhance the prompt
`enable_nsfw`	boolean	No	`true`	Allow NSFW content generation
`nsfw_check`	boolean	No	`true`	Check uploaded media inputs for NSFW content

Text-to-image and image-edit requests each generate exactly one image. The generation count is fixed internally and cannot be changed by callers.Providing image_urls changes the operation to image editing but does not switch models. This endpoint remains on xai/grok-imagine-image.

Example requests

Text to image

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "xai/grok-imagine-image",
    "input": {
      "prompt": "A robot dancing in a field of sunflowers",
      "aspect_ratio": "16:9"
    }
  }'

Image editing

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "xai/grok-imagine-image",
    "input": {
      "prompt": "Merge these images into one scene",
      "image_urls": [
        "https://example.com/img1.jpg",
        "https://example.com/img2.jpg"
      ]
    }
  }'

Response

{
  "code": 200,
  "success": true,
  "data": {
    "task_id": "abc123def456",
    "status": "completed",
    "output": {
      "image_url": "https://files.unifically.com/image/92116e95-8dd4-4cbb-b53f-ee643e1045d8.jpg"
    }
  }
}

Image-edit failure diagnostics

Before uploading edit references, the API checks Grok’s reported imageEdit quota. If the selected account explicitly has no edit quota, the request is retried with another available account. A quota response that does not expose an image-edit allowance is recorded as unknown rather than treated as exhausted. When an edit task fails, its debug_response includes diagnostics from every upstream attempt. These may include:

Field	Description
`attempt_diagnostics`	Ordered diagnostics for each attempted Grok account
`diagnosed_reason`	Machine-readable classification of the failure
`quota_precheck`	Grok’s image-edit quota status and the quota feature used for the decision
`http_status`	HTTP status returned by Grok
`event_sequence` / `event_counts`	Stream events received before the failure
`conversation_ids` / `response_ids`	Upstream identifiers useful for support investigations
`stream_errors`	Errors emitted inside Grok’s response stream
`progress_events`	Any image-generation progress events received
`raw_upstream_stream`	Complete upstream response stream for that attempt

Common diagnosed_reason values:

Reason	Meaning
`accepted_without_generation_events`	Grok accepted the edit message but emitted neither progress nor an image result
`conversation_created_without_response_events`	Grok created a conversation but did not emit response events
`quota_or_rate_limit_signal`	The response stream contained a quota or rate-limit signal
`upstream_stream_error`	Grok emitted one or more stream errors
`empty_or_unparseable_stream`	Grok returned no usable stream objects
`model_response_without_image`	A model response arrived without a generated image URL
`no_image_generation_events`	The stream completed without recognizable image-generation events

The API retains the complete upstream stream to make provider-side failures diagnosable. Avoid exposing debug_response publicly if prompts or upstream metadata may be sensitive.

Volume discounts are available for almost all models. Reach out at contact@unifically.com to discuss custom pricing.

​Model

​Parameters

​Example requests

​Text to image

​Image editing

​Response

​Image-edit failure diagnostics

Model

Parameters

Example requests

Text to image

Image editing

Response

Image-edit failure diagnostics