Grok Imagine AI: Pricing, Limits and Code (2026)

What is Grok Imagine AI?

Grok Imagine AI is xAI’s image and video generation stack, the part of Grok that turns a sentence into pictures and turns a picture into a short moving clip. It sits next to the Grok chat assistant rather than inside it: you type a prompt, you get a grid of candidate images, and you can promote any one of them into a video. The Grok apps also attach generated audio to clips, which is why the output often feels closer to a finished shot than to a silent animation test.

Two things make the family distinctive. The first is the response shape. A single text prompt returns six variations instead of one, which changes how you work: you prompt less precisely, scan the grid, and pick. The second is cost per second of video. The video models bill at a low flat rate per second, so a ten second clip is a routine experiment rather than a budget decision.

The family splits into four jobs you can call independently:

• Images from text: Grok Imagine returns six variations per prompt, or two when you attach reference images.
• Image to video: Grok Imagine Video 1.5 animates a start frame for 1 to 15 seconds.
• Text or image to video: Grok Imagine Video 1.0 covers longer shots, 6 to 30 seconds.
• Upscaling: Grok Imagine Upscale takes a finished video job and re-renders it at 720p or 1080p.

If you only ever use the phone app, that split is hidden behind one button. If you are building something, the split is the useful part, because each step has its own price and you can stop after any of them.

Is Grok Imagine AI free, and what are the limits?

This is the most common question about Grok Imagine AI, and the honest answer is that free access exists but moves. You sign in with a Grok or X account, you get a quota of generations in a rolling window, and the quota depends on your region, your platform and whether you pay for a Grok subscription. Higher quality image modes and longer videos have tended to land on the paid side first. Nobody outside xAI can give you a stable number, and any article that quotes one is quoting a snapshot.

That is the whole reason the “Grok Imagine AI is no longer free” complaint keeps resurfacing on Reddit. Usually the model did not disappear; the user hit a quota, or a mode they were used to became a subscriber feature. The practical limits of the free tier look like this:

• Rate limits you cannot see: you discover the ceiling by hitting it, mid-project.
• No programmatic access: the app is a UI. There is no key you can put in a script.
• Manual download: files come out one at a time, by hand.
• Shifting defaults: resolution and duration options change without warning.

For casual use, none of that matters and the free tier is the right choice. For anything repeatable, batch of 200 product shots, a nightly render, a feature inside your own app, you want metered pricing instead of a quota. Through the model catalog on Unicorn API, a Grok Imagine AI image call is 6 credits and credits cost $0.01 each on a monthly plan, so six images cost six cents. You can also check the exact price of any request before you run it with POST /v1/models/price, which removes the guesswork that the free tier forces on you.

The Grok Imagine AI model lineup and what each one costs

Here is the full Grok Imagine AI family as it is priced on Unicorn API. Credits are the unit: $0.01 each on a plan (Starter $25 through Pro $200), $0.015 each as a top-up that never expires.

Model	Type	What it does	Price
Grok Imagine	Image	6 variations per text prompt, 2 with reference images	6 credits per call
Grok Imagine Video 1.5	Image to video	Animate a start frame, 1-15 s, 480p/720p/1080p	3 / 4 / 7 credits per second
Grok Imagine Video 1.0	Text or image to video	6-30 s, 480p/720p/1080p	3 / 4 / 7 credits per second
Grok Imagine Upscale	Video upscale	Re-render an existing video job at higher resolution	10 credits (720p), 20 credits (1080p)

Read the image price carefully, because it is the unusual one. Six credits buys the whole grid of six images, so you pay about one credit per image, roughly a cent. Attach reference images and you still pay 6 credits but get two outputs, so each image costs three credits. The n field multiplies the price: n: 2 costs 12 credits and returns twelve images.

Video is linear in both duration and resolution, which makes budgeting trivial. A 6 second 720p clip is 24 credits, about 24 cents. The same clip at 1080p is 42 credits. Version 1.5 is image-to-video only and tops out at 15 seconds; version 1.0 accepts a text prompt alone and runs to 30 seconds, at the same per-second rate. Both can be busy at peak times, so build a retry into anything that runs unattended.

One planning note on the two video versions: the usual workflow is to generate a still with Grok Imagine AI first, pick the best of the six, then animate that exact frame with 1.5. You get far more control over composition than you do from a text-only video prompt, and the still costs a single credit.

How to call the Grok Imagine AI API

There is no public endpoint from the Grok apps, so the Grok Imagine AI API is reached through an aggregator. The pattern on Unicorn API is the same for every media model: create a job, poll it, read the output URL. Create an API key, export it as UNICORN_API_KEY, and the following works as written.

Generating six images with curl

curl -s https://api.unicornapi.net/v1/jobs \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine",
    "input": {
      "prompt": "A red fox resting in a snowy forest at dawn, soft light, highly detailed",
      "aspect_ratio": "3:2"
    }
  }'

The response carries a job id. Poll it until the status settles:

curl -s https://api.unicornapi.net/v1/jobs/JOB_ID \
  -H "Authorization: Bearer $UNICORN_API_KEY"

When status is "succeeded", the files are in outputs, and outputs[0].url is the first of the six images. Supported aspect ratios are 1:1, 2:3, 3:2, 16:9 and 9:16.

Still to video in Python

This script runs the full chain: one Grok Imagine AI image call, then an animation of the chosen frame at 720p.

import os, time, requests

BASE = "https://api.unicornapi.net"
H = {"Authorization": f"Bearer {os.environ['UNICORN_API_KEY']}"}

def run(model, payload):
    job = requests.post(f"{BASE}/v1/jobs",
                        headers=H,
                        json={"model": model, "input": payload}).json()
    job_id = job["id"]
    while True:
        job = requests.get(f"{BASE}/v1/jobs/{job_id}", headers=H).json()
        if job["status"] in ("succeeded", "failed", "canceled"):
            break
        time.sleep(3)
    if job["status"] != "succeeded":
        raise RuntimeError(f"{model} job {job_id}: {job['status']}")
    return job

images = run("grok-imagine", {
    "prompt": "A lone lighthouse on a cliff, heavy fog, cold morning light",
    "aspect_ratio": "16:9",
})
frame = images["outputs"][0]["url"]        # pick any of the six
print("picked frame:", frame)

video = run("grok-imagine-video-1-5-preview", {
    "image": frame,
    "prompt": "Slow push in, fog drifting past the lamp, waves below",
    "duration": 6,
    "resolution": "720p",
    "aspect_ratio": "16:9",
})
print("video:", video["outputs"][0]["url"])

That whole run costs 6 credits for the image grid plus 24 for the clip: 30 credits, about 30 cents on a plan. Full field reference lives in the API docs, and you can try the same prompts without writing code in the browser playground first.

Checking the price before you spend

Send the request body you are about to run to POST /v1/models/price and you get the exact charge back. Useful when duration or n is user-supplied:

curl -s https://api.unicornapi.net/v1/models/price \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"grok-imagine-video-1-5-preview",
       "input":{"duration":10,"resolution":"1080p"}}'

Upscaling a finished clip

Upscaling is a second job that references the first. Pass the task id of the video you already rendered and the resolution you want:

curl -s https://api.unicornapi.net/v1/jobs \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"grok-imagine-upscale",
       "input":{"task_id":"JOB_ID","resolution":"1080p"}}'

This is the cheap route to a 1080p deliverable. Rendering 10 seconds directly at 1080p costs 70 credits; rendering at 480p and upscaling costs 30 plus 20, so 50. The direct render usually holds more fine detail, so compare both on one shot before committing a batch.

A worked example: 40 product clips on a Starter plan

Say you need a short looping clip for each of 40 products. The Grok Imagine AI route looks like this:

1. Stills: one image call per product with the product photo as a reference image. 40 calls at 6 credits = 240 credits, and you get two options each.
2. Animation: a 4 second 720p clip from the chosen still. 40 clips at 16 credits = 640 credits.
3. Finishing: upscale the 10 clips destined for the homepage to 1080p. 10 at 20 credits = 200 credits.

Total: 1,080 credits, about $10.80 on a monthly plan. A Starter plan at $25 gives 2,500 credits, so the job uses under half of it and leaves room for the reruns you will inevitably need. Pure image work stretches much further: 2,500 credits is roughly 416 Grok Imagine AI calls, which is about 2,500 images at a credit each.

The same arithmetic tells you where to be careful. Long 1080p video is where credits go: 30 seconds at 1080p on version 1.0 is 210 credits, $2.10 for one shot. Prototype at 480p, where that shot costs 90 credits, and only commit the final cut at full resolution.

Spicy mode, NSFW and privacy on Grok Imagine AI

Two of the most searched questions about Grok Imagine AI are about adult content and privacy, so here is a plain answer to both.

Spicy mode is an age-gated setting in the consumer Grok apps. Whether you see it at all depends on your region, your platform and your subscription, and it is a toggle in app settings rather than an API parameter. Through an API provider the published content policy applies instead, and the limits there are not negotiable: generating sexual or intimate imagery of real, identifiable people without their consent is prohibited by every mainstream provider and, in a growing number of jurisdictions, by law. xAI drew heavy criticism and regulatory attention for exactly this, image editing of real people into revealing states, so expect the rules around it to keep tightening rather than relaxing.

On privacy, treat the consumer app as semi-public. Prompts and outputs can be used to improve the service, other people’s creations surface in shared feeds, and anything you post to X is public by definition. An API call has a different shape: you send a request, you receive a file URL, and nothing is published to a social feed. That makes the API the better fit for client work and unreleased products, but you should still read the data policy of whichever provider you use before sending anything confidential.

Grok Imagine AI alternatives worth keeping in the rotation

Grok Imagine AI is not the right tool for every image. Two models in the same catalog cover its weak spots, and because one key covers all of them you can switch per request rather than per project.

Model	Cost per image	Reference images	Best at
Grok Imagine	1 credit (6 per call of 6)	Yes, 2 outputs	Cheap variety, wide aspect ratios, a frame to animate
Nano Banana 2 Lite	4 credits	Up to 10	Edits and composites driven by many references
GPT Image 1.5	3 low, 4 medium, 15 high	Up to 4	Legible text, tight prompt adherence, transparent backgrounds

The decision rule is short. If you want options, use Grok Imagine AI: nothing else gives you six candidates for six credits. If you are editing an existing scene and need to feed it several references at once, Nano Banana 2 Lite accepts up to ten and costs 4 credits an image. If the image contains words that must read correctly, a poster, a packaging mockup, a UI screenshot, go to GPT Image 1.5 and pay for high quality on the final render only. For a deeper look at the image side of xAI’s stack, we covered it separately in the Grok image generator write-up, and the same job pattern applies across other video models in the catalog.

Mistakes people make with Grok Imagine AI

Most problems with Grok Imagine AI come from treating it like a one-image-per-prompt model. A few specifics worth avoiding:

• Over-engineering the prompt. You are getting six takes. A clear, concrete prompt plus selection beats a 200-word prompt plus a single render.
• Attaching references out of habit. Reference images drop the grid from six outputs to two at the same 6 credit price. Worth it for a specific product, wasteful for a mood.
• Sending a text prompt to version 1.5. It is image-to-video only. Text-only video needs version 1.0, which also covers 6 to 30 second shots.
• Polling too aggressively. A three to five second interval is plenty, and these models can be busy at peak times. Code a retry on failure rather than a tight loop.
• Not storing the task id. Upscaling needs it. Lose it and you re-render from scratch.
• Finishing at 1080p by default. Iterate at 480p for 3 credits a second and spend the 7 only on shots that made the cut.

Put together, that is the honest shape of Grok Imagine AI in 2026: a cheap, fast image model that hands you a grid instead of a guess, paired with video models whose per-second price is low enough to iterate on. The free apps are fine for trying it. The moment you need volume, repeatability, or a price you can see before you pay it, move the same models behind an API key.