xAI's video generator went from free to paywalled in March 2026, in the middle of a genuinely serious, still-unresolved safety and regulatory situation — not a minor PR headache. Here's what's factually going on, sourced, without the marketing spin.
Free image and video generation was removed on March 19, 2026. As of that date, any use of Grok Imagine requires at least a paid xAI subscription: SuperGrok Lite at $10/month (launched March 25, 2026) includes Grok Imagine capped at 15 videos per day, 480p resolution, 6-second maximum length. The full-featured tier is SuperGrok at $30/month (or $300/year), which removes those caps. There's also a $300/month "SuperGrok Heavy" tier aimed at a different, non-video use case (a multi-agent mode). These figures come from vendor pricing pages current as of August 2026 — xAI's pricing has moved before and could again, so verify current numbers before subscribing.
Grok Imagine launched in August 2025 alongside Grok 2, and included a "Spicy Mode" that reportedly produced hypersexualized images of celebrities from prompts that weren't themselves explicit. Starting in January 2026, reporting — including from NBC News and multiple outlets covering AI safety — documented that the tool could be used to generate non-consensual sexualized images of real, identifiable people, including celebrities, politicians, and reportedly minors, using ordinary photos. Those images spread widely on X.
This is not a minor content-moderation footnote. It triggered concrete regulatory and platform action:
xAI introduced technical safeguards on January 14, 2026 — but those initially applied only to non-paying users, not the paid tier. Under continued pressure, including the California AG probe, xAI later added a broader restriction blocking the editing of real people's images into revealing clothing, applied to all users. That's a real, concrete change. What it hasn't done, based on later reporting: fully solved the problem. IBTimes UK reported months after xAI's initial pledge that sexualized deepfakes of real women were still circulating. We're not in a position to independently verify the current state of that moderation — we're citing what's been reported — but the honest summary is: partial fix, contested effectiveness, ongoing investigations. Treat any claim that this is "resolved" with real skepticism and check current reporting before assuming otherwise.
On pure output quality, Grok Imagine's video model ranks respectably but not at the top of Artificial Analysis's leaderboards — around the middle of the text-to-video board and higher up (top 5) on image-to-video specifically for its newer 1.5 version, well behind the current leaders (Seedance, Gemini Omni Flash, Kling, Veo). See our full sourced rankings for exact current standings. Marketing claims that it's "the #1 AI video model" don't match Artificial Analysis's own published numbers as of this writing.
This page reports on Grok Imagine's business and regulatory situation — it isn't a how-to guide, and we're not covering prompting or usage techniques for it here.
Related: More AI video news · Current AI video model rankings · Pricing compared (Runway, Pika, Kling, Seedance, Veo)
No. Free image and video generation was removed on March 19, 2026. Any use now requires at least SuperGrok Lite ($10/month, launched March 25, 2026), which includes Grok Imagine limited to 15 videos per day at 480p resolution and 6-second maximum length. Full access requires the $30/month SuperGrok tier.
Starting in January 2026, reporting found that Grok could be used to generate non-consensual sexualized images of real, identifiable people — including celebrities, politicians, and reportedly minors — from ordinary photos, which spread widely on X. This triggered investigations by the European Commission and Ireland's Data Protection Commission (under GDPR), bans in Malaysia and Indonesia, criticism from UK regulators and the Prime Minister's office, and a probe from California's attorney general.
Yes, according to reporting from April 2026, Apple threatened to remove Grok from the App Store specifically over the sexualized deepfake issue. xAI has since added restrictions, including blocking the editing of real people's images into revealing clothing, though independent reports later in 2026 indicated the underlying problem was not fully resolved.
Partially, and unevenly. xAI introduced technical safeguards on January 14, 2026, but those initially applied only to non-paying users. It later added a broader restriction blocking edits of real people into revealing clothing for all users, announced amid the California attorney general's probe. Independent reporting later in 2026, including from IBTimes UK, found sexualized deepfakes of real women still circulating months after xAI's initial pledge to address the issue — so treat any claim that it's fully resolved with skepticism and check current reporting.