Grok Imagine Video Generation

Grok Imagine (grok-imagine-video series) is xAI's video generation model, served through the unified OpenAI-style /v1/videos endpoint. The platform transparently translates Sora-style parameters into xAI official fields (e.g. size → resolution), so both styles work out of the box.

Models

• grok-imagine-video: standard tier, text/image/video input;
• grok-imagine-video-1.5: latest tier with native audio output, up to 1080p;
• grok-imagine-video-1.5-preview: 1.5 preview, priced the same as the GA model.

Endpoints

• Create task: POST $URL/v1/videos
• Query task: GET $URL/v1/videos/{task_id}
• Download video: GET $URL/v1/videos/{task_id}/content

Supported Models

Model	Positioning	Resolutions	Input
grok-imagine-video	Standard	480p / 720p	text / image / video
grok-imagine-video-1.5	Latest (native audio)	480p / 720p / 1080p	text / image / audio
grok-imagine-video-1.5-preview	1.5 preview (same price)	480p / 720p / 1080p	text / image / audio

Parameters

Field	Type	Required	Description
model	string	Yes	One of the models above
prompt	string	Yes	Video description (optional for image-to-video)
seconds	string | number	No	Duration in seconds, 1–15, default 8; duration (integer) is equivalent
resolution	string	No	xAI tiers: 480p / 720p (default) / 1080p
size	string	No	Sora/OpenAI style: 480p / 1280x720 / 720x1280 etc. — translated to resolution and removed automatically (ignored if resolution is also set)
image_url	string	No	Image-to-video: reference image URL or base64 data URL; Sora-style image / input_reference are auto-normalized to image_url
reference_images	array	No	Multi-image reference: [{"url": "..."}] at the top level, up to 7 images; reference each one in the prompt with <IMAGE_1>…<IMAGE_N> tags. Supported on all three models; a single image can also use image_url, and both may coexist
aspect_ratio	string	No	16:9 / 9:16 / 1:1 etc. (passed through to xAI)

Constraints

• Durations outside 1–15 seconds are rejected with 400 before any billing occurs;
• 4K is not supported: size values 2k / 4k are blocked; oversized pixel sizes fall back to 1080p.

Request Example

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-video-1.5",
    "prompt": "A lantern floating into the night sky",
    "seconds": "4",
    "size": "480p"
  }'

Image-to-Video (image_url)

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-video-1.5",
    "prompt": "Bring the scene to life",
    "seconds": "4",
    "resolution": "720p",
    "image_url": "https://example.com/photo.jpg"
  }'

Multi-Image Reference (reference_images)

Pass multiple reference images in the top-level reference_images array (up to 7) and reference each one in the prompt with <IMAGE_N> tags:

curl $URL/v1/videos \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-video-1.5",
    "prompt": "The character in <IMAGE_1> slowly turns their head, with the scene from <IMAGE_2> as the background. Slow push-in.",
    "seconds": "4",
    "resolution": "720p",
    "reference_images": [
      { "url": "https://example.com/character.jpg" },
      { "url": "https://example.com/scene.jpg" }
    ]
  }'

Notes

• reference_images works identically on grok-imagine-video / grok-imagine-video-1.5 / grok-imagine-video-1.5-preview — just change the model name;
• When multiple images are passed via the Sora-style unified entry images / input_reference, the platform automatically merges images 2+ into reference_images, and billing follows the actual count;
• Each reference image is billed at the "Input image" rate above (frozen at submit time).

Query and Download

# queued / processing / completed / failed
curl $URL/v1/videos/{task_id} \
  -H "Authorization: Bearer $YOUR_API_KEY"

curl -o video.mp4 $URL/v1/videos/{task_id}/content \
  -H "Authorization: Bearer $YOUR_API_KEY"

Pricing

Billed by output seconds × resolution, plus official media input fees for reference images/videos (frozen at submit time):

Model	480p	720p	1080p	Input image	Input video
grok-imagine-video	$0.05/s	$0.07/s	—	$0.002/image	$0.01/s
grok-imagine-video-1.5	$0.08/s	$0.14/s	$0.25/s	$0.01/image	—
grok-imagine-video-1.5-preview	$0.08/s	$0.14/s	$0.25/s	$0.01/image	—

Example: grok-imagine-video-1.5, 4 s at 480p = 4 × $0.08 = $0.32; with 3 reference images add 3 × $0.01 = $0.03, for a total of $0.35.

Notes

• Videos expire upstream: downloading an expired artifact returns 410 Gone — download promptly after completion;
• Default duration is 8 seconds (xAI default) — requests without seconds/duration are billed for 8 s;
• Task statuses: queued → processing (0–99%) → completed / failed;
• Group multipliers apply on top of the base prices above.