Grok AI Video Generator: Cost per Second and API (2026)

The Grok AI video generator is xAI’s Grok Imagine model. You give it a still image, a text prompt or both, and it returns a short animated clip. It is fast, its motion holds up well, and per second of video it is one of the cheapest models you can call. This post covers how it works, what free use gets you, what each clip costs through an API, working code, and prompt patterns that save you from wasted generations.

What is the Grok AI video generator?

Grok is the AI assistant from xAI, launched in November 2023. Imagine is the part of Grok that makes images and video. When people search for a Grok AI video generator, they usually mean one of two things: the Imagine feature inside the Grok app and website, or the Grok Imagine Video model underneath it, which developers can call from their own code.

The model works in two modes:

• Text-to-video: you describe a scene in words and the model creates everything, including the subject, setting, lighting and motion.
• Image-to-video: you supply a start frame, which is the image that becomes the first frame of the clip, and the model animates it. The subject, composition and colours stay close to your source. That makes this mode the right choice for product shots, characters and brand assets you have already approved.

Grok Imagine Video 1.0 and 1.5

Two versions are available, and they suit different jobs:

• Grok Imagine Video 1.5 (released 2026-06-01) is image-to-video only. It animates a start frame at 480p, 720p or 1080p, for anywhere from 1 to 15 seconds. Its aspect ratios are auto, 1:1, 16:9, 9:16, 3:2 and 2:3. It has strong motion quality, but it can be busy at peak times.
• Grok Imagine Video 1.0 (released 2025-11-03) handles both text-to-video and image-to-video. It produces clips of 6 to 30 seconds at the same three resolutions, and it costs the same per second as 1.5.

The rule of thumb is simple. If you are starting from a picture and need 15 seconds or less, use 1.5. If you only have a text prompt, or you need a clip longer than 15 seconds, use 1.0.

Is the Grok AI video generator free?

You can try the Grok AI video generator in the Grok app without paying, but free use is capped. xAI adjusts those caps often, so any specific number you find in a forum post or video may already be out of date. Heavier use requires a paid Grok subscription. For the current limits, check the app itself rather than a third-party summary.

Searches like “free unlimited” and “free apk” deserve some caution. No provider offers unlimited free video generation, because every clip costs real GPU time (GPUs are the graphics processors that run these models). Sites that promise unlimited free Grok video are at best a wrapper with hidden limits of their own. Unofficial APKs can be malware or tools for stealing account logins. Get the official app from the official app store.

That also answers the broader question of which AI video generator is 100% free: none is, if you use it regularly. Grok, Canva, CapCut and others offer free tiers that are fine for experiments. Those tiers come with daily caps, watermarks, lower resolution or slower queues. If you need dozens of clips a week, a predictable per-clip price is less frustrating than rotating between free accounts.

App or API?

• The Grok app is interactive and quick to start. It suits personal projects and one-off clips, and it is billed by subscription with usage limits.
• An API charges per clip and has no daily cap beyond your credit balance. You can script batches, plug generation into a product, and switch to another model when Grok is busy.

How much does the Grok AI video generator cost per clip?

Through Unicorn API, both Grok versions are priced per second of output: 2.4 credits at 480p, 4.5 credits at 720p and 8 credits at 1080p. Monthly plans give you 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. Each output is rounded up to whole credits, so a 1-second 480p test costs 3 credits, not 2.4.

Clip	Model	Credits	On a plan	With top-ups
5 s, 480p	1.5	12	$0.12	$0.18
6 s, 720p	1.0 or 1.5	27	$0.27	about $0.41
10 s, 480p	1.0 or 1.5	24	$0.24	$0.36
10 s, 720p	1.0 or 1.5	45	$0.45	about $0.68
10 s, 1080p	1.0 or 1.5	80	$0.80	$1.20
15 s, 1080p	1.0 or 1.5	120	$1.20	$1.80
30 s, 1080p	1.0 only	240	$2.40	$3.60

For a budget check, take the Creator plan. Its 5,000 credits cover about 111 ten-second clips at 720p, or 62 at 1080p. The bigger saving comes from how you work: 480p is about 3x cheaper per second than 1080p. Do your drafts at 480p, and only spend 1080p credits on the take you will actually publish.

You never have to guess the price. Send the model and input to POST /v1/models/price and you get the exact credit cost before anything runs. Full plan details are on the pricing page.

How to use the Grok AI video generator with an API

Video generation is asynchronous: you submit a job, the model renders it, and you collect the result. With Unicorn API the flow is the same for every video model:

1. Create an API key on the keys page and export it as UNICORN_API_KEY.
2. Put your start image at a public URL. Grok Imagine Video 1.5 needs one, because it is image-to-video only.
3. Send POST /v1/jobs with the model id and an input object.
4. Poll GET /v1/jobs/{id} until status is "succeeded".
5. Download the video from outputs[0].url and save it to your own storage.

Submit an image-to-video job 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/sneaker.jpg",
      "prompt": "The sneaker rotates slowly on a turntable, soft studio light sweeps across the leather, camera dollies in gently",
      "duration": 6,
      "resolution": "720p",
      "aspect_ratio": "9:16"
    }
  }'

The response contains a job id. That request costs 27 credits.

A complete Python script with polling

import os
import time
import requests

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


def generate_video(body, poll_every=5, timeout=900):
    r = requests.post(f"{API}/v1/jobs", json=body, headers=HEADERS, timeout=30)
    r.raise_for_status()
    job = r.json()

    deadline = time.time() + timeout
    while job["status"] != "succeeded":
        if job["status"] == "failed":
            raise RuntimeError(f"Job {job['id']} failed: {job}")
        if time.time() > deadline:
            raise TimeoutError(f"Job {job['id']} still running after {timeout}s")
        time.sleep(poll_every)
        r = requests.get(f"{API}/v1/jobs/{job['id']}", headers=HEADERS, timeout=30)
        r.raise_for_status()
        job = r.json()

    return job["outputs"][0]["url"]


def download(url, path):
    with requests.get(url, stream=True, timeout=120) as r:
        r.raise_for_status()
        with open(path, "wb") as f:
            for chunk in r.iter_content(chunk_size=1 << 20):
                f.write(chunk)


# Image-to-video with Grok Imagine Video 1.5
url = generate_video({
    "model": "grok-imagine-video-1-5-preview",
    "input": {
        "image": "https://example.com/sneaker.jpg",
        "prompt": "The sneaker rotates slowly on a turntable, soft studio light "
                  "sweeps across the leather, camera dollies in gently",
        "duration": 6,
        "resolution": "720p",
        "aspect_ratio": "9:16",
    },
})
download(url, "sneaker.mp4")

# Text-to-video with Grok Imagine Video 1.0 (6 to 30 seconds)
url = generate_video({
    "model": "grok-imagine/image-to-video",
    "input": {
        "prompt": "A slow drone shot over a rocky coastline at sunset, waves breaking",
        "duration": 10,
        "resolution": "480p",
        "aspect_ratio": "16:9",
    },
})
download(url, "coast.mp4")

The text-only call goes to version 1.0, because 1.5 always needs an image. Every field and model id is listed in the API documentation. If you would rather test prompts by hand first, the video playground runs the same models in the browser. If you work in Cursor or Claude Code, you can also connect the API to your agent over MCP and ask it to generate clips directly.

How to write Grok AI video generator prompts

Most weak results come from prompts that describe a picture instead of a movement. A Grok AI video generator prompt works best when it answers four questions, in this order:

1. What moves? Name the subject and one clear action: “the dancer spins once”, “steam rises from the cup”.
2. How does the camera move? Use film terms the model understands: dolly in, pan left, orbit, static tripod shot, handheld, drone shot.
3. What is the light and setting? For example “golden hour”, “soft studio light”, “neon reflections on wet asphalt”.
4. What is the pace and style? For example “slow motion”, “calm and steady”, “cinematic, shallow depth of field”.

For image-to-video, skip the subject description. The model can already see your start frame, and describing it again only invites the model to redraw it. Spend those words on motion and camera instead.

Worked example: a product clip for social

Say you have a clean product photo of a sneaker and want a 6-second vertical clip.

• Weak prompt: “A cool white sneaker, make it look awesome.” This gives the model no motion, no camera and no light, so it guesses, and every attempt comes back different.
• Strong prompt: “The sneaker rotates slowly on a turntable, soft studio light sweeps across the leather, camera dollies in gently, calm pace, shallow depth of field.”

The budget for this job looks like this. Run three variations at 480p for 6 seconds, at 15 credits each after rounding. Pick the best one, then rerun that prompt at 1080p for 48 credits. The total is 93 credits, or $0.93 on a monthly plan. Doing all four runs at 1080p would cost 192 credits for the same final clip.

Three more habits help. Use the auto aspect ratio on 1.5 when your start image already has the framing you want. Keep loops short: 1.5 accepts durations as short as 1 second, which suits looping GIF-style clips. Change one variable per run, so you can tell which part of the prompt made the difference.

Grok vs Canva, CapCut, Sora and other AI video generators

Many tools called “AI video generators” are actually editors with AI features, not generative models. Knowing which category you need saves a lot of trial and error.

Tool	What it is	Best for
Grok Imagine	Generative video model from xAI	Short cinematic clips, animating stills, low cost per second
Canva’s AI video generator	AI clips inside the Canva design editor	Social posts and presentations you already build in Canva
CapCut AI video generator	Video editing app with AI features	Short-form editing, captions, templates
InVideo AI video generator	Turns a script into an assembled video with stock footage and voiceover	Explainers and faceless channel content
Adobe Firefly AI video generator	Adobe’s generative video, tied to its creative apps	Teams already working in Adobe tools
Synthesia AI video generator	AI avatar presenter videos	Training and talking-head content
Sora AI video generator	OpenAI’s text-to-video model	OpenAI wound down the Sora brand in 2026, so check current availability before you build on it

If you need raw generated footage for your own editor or app, compare Grok with other generative models rather than with editors. These are 10-second 720p clips on Unicorn API:

Model	Modes	10 s at 720p	Notes
Grok Imagine Video 1.5 / 1.0	Image-to-video (1.5); text or image (1.0)	45 credits	Up to 15 s (1.5) or 30 s (1.0)
Wan 2.6 Flash Image To Video	Image-to-video	40 credits silent, 80 with audio	Fixed 5, 10 or 15 s options
Vidu Q4 Image To Video	Image-to-video	152 credits	Goes up to 2k and 4k
Wan 3.0	Text or image	160 credits	Native synced audio at no extra cost, up to 30 s
Wan 3.0 Video Prime	Text, first and last frame, references	252 credits	Most control over references

Grok wins on price per second. Wan 3.0 is the simple pick when you need sound generated with the picture. Vidu Q4 is the pick when you need resolutions above 1080p. For a deeper comparison, see our Grok Imagine alternatives guide, and for other approaches to animating stills, read how Kling handles image-to-video. The full model catalog lists every option under one key.

Common mistakes with the Grok AI video generator

• Sending a text-only prompt to 1.5. Version 1.5 is image-to-video only. For text-to-video, use 1.0, or generate a start image first and pass it in.
• Requesting durations outside the supported range. Version 1.5 accepts 1 to 15 seconds and 1.0 accepts 6 to 30. Combinations outside the price table are billed at a flat default (22.5 credits on 1.5, 27 on 1.0), so validate inputs in your code before you submit.
• Drafting at 1080p. Iterate at 480p and upgrade only the winning take. That one habit cuts draft costs by about 3x.
• Fighting the start image’s framing. If you ask for 16:9 output from a tall portrait photo, the model has to invent the missing edges. Match the aspect ratio to your source, or use auto on 1.5.
• Describing the image instead of the motion. The model already sees the subject. Tell it what moves and how the camera behaves.
• Polling without a timeout. Version 1.5 can be busy at peak times. Set a deadline, as the script above does. If a batch is time-sensitive, fall back to 1.0 or a Wan model rather than waiting.
• Hotlinking output URLs. Download finished clips to your own storage before you publish or reuse them.
• Trusting “free unlimited” offers. They hide limits, watermarks or worse. A known per-clip price is cheaper than a broken workflow.

Use the Grok app to explore ideas, and once you need volume, call the Grok AI video generator through an API with a clear price on every clip. Create a key, run one 480p draft, and you will know within a minute whether the motion suits your project.