The News: Elon Musk confirmed back-to-back improvements to both Grok Imagine (image creation) and Grok Video, signaling a rapid development push at xAI.
Why It Matters: xAI is accelerating its multimodal AI capabilities at a pace that directly affects how Tesla owners, developers, and X Premium subscribers interact with Grok — from generating marketing visuals to creating short AI videos from text prompts.
Source: @elonmusk on X | Grok Video post
Two Announcements, Eleven Minutes Apart
In a rapid-fire pair of posts early Thursday morning, Elon Musk signaled that xAI is not slowing down on its generative AI ambitions. First came a quiet but loaded update — "Grok Imagine improves again" — followed just eleven minutes later by a two-word declaration: "Grok Video 🥇". The trophy emoji wasn't accidental. xAI appears to be positioning Grok Video as a best-in-class contender in the AI video generation space.
📊 Key Figures
| Metric | Value | Context |
|---|---|---|
| Video Resolution | 720p | Max supported |
| Base Clip Length | 6–10 sec | Per generation |
| Max Chained Length | 15 sec | Via Extend from Frame |
| API Price (720p + audio) | $0.05/sec | ~$0.50 per 10-sec clip |
| Training Infrastructure | 110,000 GPUs | NVIDIA GB200 cluster |
| Grok Imagine 1.0 Launch | Feb 3, 2026 | 720p + improved audio |
| API Launch Date | Jan 28, 2026 | Text-to-video, img-to-video |
| Access Tier | X Premium | Required for all video features |
What's Actually New: Grok Imagine vs. Grok Video
These aren't two separate products — they're two faces of the same generative engine. Here's how to think about the distinction:
Grok Imagine is xAI's image and video generation suite. The "improves again" announcement follows a consistent pattern: Grok Imagine 1.0 launched February 3, 2026 with 720p support and improved audio. On March 2, xAI added the "Extend from Frame" feature — letting users chain video clips by using the final frame of one generation as the starting point of the next. The result is smoother continuations that preserve motion, character positioning, and lighting. Today's update appears to push quality further, though xAI has not yet published detailed release notes.
Grok Video is the output format — the AI-generated video clips themselves, powered under the hood by xAI's proprietary Aurora autoregressive engine, trained on a cluster of 110,000 NVIDIA GB200 GPUs. The 🥇 emoji from Musk suggests xAI believes its video quality now leads the field, at least on certain benchmarks or visual quality metrics.
One caveat worth noting: community testing in March 2026 has confirmed that video quality visibly degrades after multiple chained extensions. xAI has not provided a timeline for a fix. That's a real limitation for anyone trying to produce longer-form content.
Organization Gets Easier: Folders Are Now Live
Alongside the quality improvements, xAI quietly rolled out a folders feature within Grok Imagine on March 4, 2026 — one day before today's announcements. Users can now organize generated images and videos into named folders, a quality-of-life upgrade that matters more as the volume of AI-generated content grows. It's a small feature, but it signals that xAI is thinking about Grok Imagine as a creative workspace, not just a demo tool.
Who Can Access This Right Now
Access to all video generation features — including Extend from Frame and today's improvements — requires an X Premium subscription. Grok Imagine was initially made available to SuperGrok and Premium+ subscribers on iOS in August 2025, with broader rollout following. Developers who want programmatic access can use the Grok Imagine API at api.x.ai, which launched January 28, 2026, supporting text-to-video, image-to-video, and video editing workflows.
API pricing: $0.05 per second for 720p video with audio — or roughly $0.50 for a 10-second clip, which works out to $4.20 per minute of generated video. For developers building applications, that's a competitive price point relative to other video generation APIs currently on the market.
🔭 The BASENOR Take
Timeline: Rapid iteration — major updates every 2–4 weeks since January 2026
Impact Level: Medium-High for X Premium subscribers and developers; Low for Tesla vehicle owners directly
Confidence: High — multiple verified sources corroborate the technical specs and pricing
The speed of xAI's iteration here is the real story. From API launch (January 28) to Grok Imagine 1.0 (February 3) to Extend from Frame (March 2) to today's quality push — that's four meaningful updates in five weeks. The Aurora engine running on 110,000 GB200 GPUs gives xAI a compute foundation that few competitors can match for training and inference at scale.
For Tesla owners specifically, the connection isn't direct today — but it's not invisible either. The same AI infrastructure and engineering culture driving Grok's rapid improvement is the same one feeding into Tesla's FSD and Optimus development pipelines. xAI and Tesla share Elon Musk's strategic attention and, increasingly, technical talent and compute resources. A Grok that gets dramatically better at understanding and generating video is also a Grok that gets better at the kind of world-model reasoning that autonomous driving depends on.
The 15-second clip ceiling and quality degradation on chained extensions are genuine limitations today. But given the pace of updates, those constraints look more like version 1.x growing pains than fundamental architectural problems. Watch for a resolution bump beyond 720p and longer native clip lengths as the next likely milestones.
The 🥇 from Musk is a bold claim. Whether Grok Video actually leads the field on objective benchmarks remains to be independently verified. But the trajectory — and the compute behind it — makes the ambition credible.





![BASENOR Phone Mount for 2025 2026 Tesla Model Y Juniper/Model 3 Highland, Dashboard Phone Holder Does Not Block View [No Adhesive][Dual Arms][360° Adjustable] Tesla Accessories Fit All Smartphone](http://www.basenor.com/cdn/shop/files/basenor-phone-mount-for-2025-2026-tesla-model-y-juniper-model-3-highland.jpg?v=1768393169&width=400)


