Grok AI Image to Video: Prompts, Pricing and API (2026)

Grok AI image to video is the feature that takes one still picture and turns it into a short moving clip using xAI’s Grok Imagine model. You give it a start frame and a sentence about what should happen, and a few moments later you get a video in which the subject moves, the camera drifts and the light shifts. This guide covers how it works, which version to pick, what each clip costs down to the credit, how to write prompts that behave, and how to run it from code instead of tapping through the app.

What is Grok AI image to video?

Grok is the assistant built by xAI, and Grok Imagine is its family of image and video generation models. In the Grok app and on X, you can generate a picture, then press a button to animate it. The same models are also available over an API, which is how developers put Grok AI image to video inside their own products, batch jobs and content pipelines.

Under the hood, these are diffusion models. A diffusion model learns to build an image (or a sequence of frames) by starting from random noise and removing it step by step until the result matches the input. For image-to-video, the first frame is pinned to your picture, so the model’s job is to invent plausible motion that starts from exactly that image. This is different from text-to-video, where the model invents the scene as well as the motion.

Two practical consequences follow from that design:

• Your image decides the look. Lighting, composition, faces and products come from the start frame. If the picture is soft, cluttered or badly cropped, the video will be too.
• Your prompt decides the motion. The model already sees the picture. Describing it again wastes words; describing what moves is what changes the result.

There are two routes to the feature. The consumer route is the Grok app, which is quick for one-off clips but comes with usage limits and manual downloads. The developer route is an API, where you send a request with an image URL and settings, and get back a video file. Unicorn API offers both current Grok Imagine versions behind one key, alongside the other major video models, so you can test Grok against alternatives without new accounts.

Grok Imagine 1.0 vs 1.5: which version should you use?

Two Grok Imagine video models are available, and they overlap more than their names suggest. Both cost the same per second at each resolution. The differences are clip length and input type.

	Grok Imagine Video 1.0	Grok Imagine Video 1.5
Model id	grok-imagine/image-to-video	grok-imagine-video-1-5-preview
Inputs	Text-to-video or image-to-video	Image-to-video only
Duration	6 to 30 seconds	1 to 15 seconds
Resolutions	480p, 720p, 1080p	480p, 720p, 1080p
Aspect ratios	1:1, 16:9, 9:16, 3:2, 2:3	auto, 1:1, 16:9, 9:16, 3:2, 2:3
Per-second rate	2.4 / 4.5 / 8 credits	2.4 / 4.5 / 8 credits
Notes	Longer clips	Strong motion quality; can be busy at peak times

A simple rule works for most people:

• Use 1.5 by default for Grok AI image to video work under 15 seconds. It accepts very short clips (1 to 5 seconds) that 1.0 cannot produce, which matters for social loops, product spins and UI previews. The auto aspect ratio also lets it match your source image.
• Use 1.0 when you need 16 to 30 seconds in one take, when you want text-to-video with no start image, or when 1.5 is congested and you need a result now.

The “busy at peak times” note on 1.5 is worth planning for. If you run a production queue, build a fallback: try 1.5 first and route to 1.0 (at 6 seconds or more) when jobs stall. Since the rates match, the fallback does not change your costs for the same length and resolution.

How much does Grok AI image to video cost?

On Unicorn API every request is priced in credits. Monthly plans give credits at $0.01 each (Starter $25, Creator $50, Studio $100, Pro $200 per month). Top-up credits cost $0.015 each and never expire. Full details are on the pricing page.

Grok Imagine charges per clip, and the clip price is the per-second rate times the duration: 2.4 credits per second at 480p, 4.5 at 720p and 8 at 1080p. The total is rounded up to whole credits, so a 1-second 480p clip (2.4) is billed as 3 credits.

Clip	Model	Credits	Plan price	Top-up price
3 s, 480p	1.5	7.2 → 8	$0.08	$0.12
5 s, 720p	1.5	22.5 → 23	$0.23	$0.345
6 s, 480p	1.0 or 1.5	14.4 → 15	$0.15	$0.225
10 s, 720p	1.0 or 1.5	45	$0.45	$0.675
15 s, 1080p	1.0 or 1.5	120	$1.20	$1.80
30 s, 1080p	1.0	240	$2.40	$3.60

To put that in project terms: the Starter plan’s 2,500 credits cover about 108 five-second 720p clips, or about 166 six-second 480p drafts. Because resolution multiplies cost (1080p is about 3.3x the price of 480p), the cheapest workflow is to draft at 480p, pick the takes you like, and re-render only those at 1080p with the same prompt and image.

You never have to guess. Send the same model and input to POST /v1/models/price and you get the exact cost before anything runs. That is useful when you let users choose duration and resolution in your own app and want to show the price up front.

How to run Grok AI image to video through the API

Video generation takes longer than a normal web request, so it runs as a job. You create the job, then check on it until it finishes. The flow has four parts:

1. Create an API key on the keys page and store it as UNICORN_API_KEY.
2. Host your start image at a public URL (your own storage bucket or CDN is fine).
3. Send POST /v1/jobs with the model id and an input object.
4. Poll GET /v1/jobs/{id} until status is "succeeded", then read outputs[0].url.

The Grok Imagine models accept these input fields: mode, image, prompt, duration, resolution and aspect_ratio. The accepted values for mode are listed in the API documentation; the example below leaves it unset.

Quick test with curl

curl -X POST https://api.unicornapi.net/v1/jobs \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-video-1-5-preview",
    "input": {
      "image": "https://example.com/coffee-cup.jpg",
      "prompt": "Steam rises slowly from the cup, slow push-in, soft morning light flickers across the table",
      "duration": 5,
      "resolution": "720p",
      "aspect_ratio": "auto"
    }
  }'

# Then, using the id from the response:
curl https://api.unicornapi.net/v1/jobs/JOB_ID \
  -H "Authorization: Bearer $UNICORN_API_KEY"

Complete Python example with price check and polling

import os
import time
import requests

API = "https://api.unicornapi.net"
HEADERS = {
    "Authorization": f"Bearer {os.environ['UNICORN_API_KEY']}",
    "Content-Type": "application/json",
}

body = {
    "model": "grok-imagine-video-1-5-preview",
    "input": {
        "image": "https://example.com/coffee-cup.jpg",
        "prompt": (
            "Steam rises slowly from the cup, slow push-in, "
            "soft morning light flickers across the table"
        ),
        "duration": 5,
        "resolution": "720p",
        "aspect_ratio": "auto",
    },
}

# 1. Check the exact price before running anything
price = requests.post(f"{API}/v1/models/price", json=body, headers=HEADERS)
price.raise_for_status()
print("Price:", price.json())

# 2. Create the job
resp = requests.post(f"{API}/v1/jobs", json=body, headers=HEADERS)
resp.raise_for_status()
job_id = resp.json()["id"]
print("Job created:", job_id)

# 3. Poll until it finishes (give up after 10 minutes)
deadline = time.time() + 600
while time.time() < deadline:
    job = requests.get(f"{API}/v1/jobs/{job_id}", headers=HEADERS).json()
    status = job["status"]
    if status == "succeeded":
        print("Video:", job["outputs"][0]["url"])
        break
    if status == "failed":
        raise RuntimeError(f"Job failed: {job}")
    time.sleep(5)
else:
    raise TimeoutError("Job did not finish in time")

To switch to the longer model, change model to grok-imagine/image-to-video and set duration between 6 and 30. Everything else stays the same. If you would rather click than code for your first few tests, the video playground runs the same models in the browser, and if you work inside Cursor or Claude Code, the MCP setup lets your agent submit jobs for you.

How to write Grok image to video prompts that work

Most disappointing results come from prompts that describe the picture instead of the motion. The model can already see a red car on a wet street; telling it so adds nothing. Telling it the car pulls away and the camera pans to follow gives it something to do.

A pattern that holds up across Grok AI image to video clips is subject action, camera move, atmosphere:

• Subject action: “the dog shakes water off its coat”
• Camera move: “slow push-in”, “handheld tracking shot”, “static locked-off camera”, “orbit left around the subject”
• Atmosphere: “dust drifts in the sunbeam”, “neon signs flicker”, “leaves rustle in a light wind”

Some practical rules:

• Match the action count to the length. One action per 3 to 5 seconds is a sensible ceiling. Ask for five things in a 5-second clip and the model will rush or skip them.
• Say what should stay still. For product shots, “the bottle stays fixed, only the camera moves” reduces warping of labels and logos.
• Use pacing words. “Slowly”, “gently” and “gradually” calm the motion. Without them, models tend to add energetic movement.
• Pick the aspect ratio before you generate the image. If the clip is for vertical social video, start from a vertical image and use 9:16. Forcing a landscape photo into a vertical frame crops or stretches it.
• Keep faces large and well lit in the start frame. Small, shadowed faces are where motion artefacts show up first.

A worked example

Say you sell ceramic mugs and want a 5-second vertical clip for each product. Your start frame is a studio photo of a mug on a wooden table, shot vertically. A weak prompt is “a beautiful ceramic mug on a table”. A strong one is: “Hot coffee steams gently from the mug, the camera slowly pushes in, warm window light shifts across the table, the mug stays perfectly still.”

Draft it at 480p for 5 seconds on Grok Imagine 1.5 (12 credits). Run three takes for 36 credits, keep the best, and re-render that one at 1080p (40 credits). Total: 76 credits, or $0.76 on a monthly plan, for a finished clip you chose from three options. Across a catalogue of 30 mugs, that is 2,280 credits, which fits inside the Starter plan.

Grok vs Kling, PixVerse, Vidu and Wan for image to video

Grok is not the only image-to-video model worth knowing, and one of the advantages of a single API is that switching costs you a model id, not a new account. Here is how the main options compare for a 5-second 720p clip without audio, which is a common starting point.

Model	5 s, 720p cost	Audio option	What stands out
Grok Imagine Video 1.5	23 credits	No audio input	Very low per-second cost, strong motion, 1 to 15 s
Wan 2.6 Flash Image To Video	20 credits	Yes (40 credits with audio)	Priced at 5, 10 and 15 s; multi-shot input
PixVerse V6 Max Image To Video	44 credits	Yes (60 credits with audio)	Low-cost 360p drafts at 4.8 credits per second
Vidu Q4 Image To Video	76 credits	Yes	Goes up to 2K and 4K output
Kling 4.0 Keyframe To Video	143 credits	Not listed	Set first frame, last frame and keyframes

How to read this:

• For volume and cheap iteration, Grok Imagine 1.5 and Wan 2.6 Flash are the two to beat. Grok gives you any length from 1 to 15 seconds; Wan is priced in 5-second steps.
• When you need sound in the same request, Grok’s inputs have no audio switch, so pick Wan, PixVerse or Vidu, or add audio afterwards.
• When the ending matters, Kling 4.0 Keyframe lets you pin the last frame as well as the first. That is the tool for transitions, before-and-after shots and loops that must land exactly. People searching for Kling AI image to video usually want this kind of control.
• When you need consistent characters or products across shots, the reference models help: Kling 4.0 Reference To Video accepts reference images and videos, and Vidu Q4 Reference To Video accepts reference images and audio.
• For the highest resolution, Vidu Q4 is the only one here that offers 2K and 4K.

Design tools such as Canva also offer image-to-video inside their editors, which is convenient for a one-off social post. The trade-off is that you work through an interface, one clip at a time. If you need dozens of clips, consistent settings or automation, an API is the more practical path. You can browse every option in the model catalog.

Is there a free, unlimited Grok AI image to video option?

This is the most searched question around the topic, so here is the plain answer: no service offers truly unlimited free video generation for long. Each clip uses expensive GPU time, so “free” tiers come with daily caps, queues, watermarks or limits that change without much notice. The Grok app has offered some free Grok AI image to video generations, but you should treat that as a trial, not a production plan.

Many sites that promise free unlimited Grok video are resellers wrapping someone else’s API, often with unclear terms about what happens to your uploaded images. If your start frames are client photos or unreleased products, that matters.

The honest low-cost route is paying per clip at a rate you can see in advance. At 2.4 credits per second at 480p, a 5-second Grok draft costs $0.12 on a monthly plan. Top-up credits never expire, so occasional users are not paying for a subscription they do not use.

Common mistakes with Grok AI image to video

• Rendering everything at 1080p. Drafting at 480p and upgrading only the winners cuts cost by about 3x on discarded takes.
• Asking 1.0 for short clips. Grok Imagine 1.0 starts at 6 seconds. For 1 to 5 seconds, use 1.5.
• Using a private image URL. The model must be able to fetch your start frame. Signed URLs that expire in seconds, or files behind a login, will fail.
• Re-describing the image. Spend the prompt on motion and camera, as covered above.
• Polling with no timeout. Jobs can queue, especially on 1.5 at peak times. Always set a deadline and a fallback model.
• Expecting long single takes. For anything past 30 seconds, generate clips in sequence and use the last frame of one as the start image of the next, then join them in an editor.
• Forgetting rounding. Odd-second clips at 480p and 720p land on fractional credits and round up. It is small per clip but adds up across thousands of jobs, which is another reason to check prices with the price endpoint.

Get these right and Grok AI image to video becomes a dependable, inexpensive part of a pipeline rather than a novelty in an app. Start with a 480p test in the playground, settle your prompt pattern, then move the same request into code.