Grok Imagine After Sora: xAI's 2026 Video Push

Update, 2 Nov 2026: This article was first published on 27 March 2026, days after OpenAI said it would close Sora. It has been rewritten with the confirmed shutdown dates and the xAI video releases that followed, and speculation about the promised update has been removed.

Grok Imagine, the image and short-video generator from Elon Musk’s xAI, kept growing after OpenAI walked away from consumer AI video. OpenAI announced on 24 March 2026 that it would discontinue Sora; within a day Musk posted that a “big” Grok Imagine update was coming and that xAI was doubling down on video.

Seven months on, the record is clearer. OpenAI’s help centre says the Sora web and app experiences ended on 26 April 2026 and the Sora API on 24 September 2026 [1]. xAI, meanwhile, shipped Grok Imagine Video 1.5 in June and a cheaper Lite model in October [2][3]. This article sets out what each company actually did, and what it means for people who make short videos.

How OpenAI shut down Sora in 2026

Sora was first shown in February 2024 and later relaunched as a social app built around the Sora 2 model. Reuters reported on 24 March 2026 that the decision to drop it startled Disney, which had been working with OpenAI [4]. OpenAI has not published a detailed public explanation of why it closed the product.

The practical steps are set out in OpenAI’s own support page [1]:

The Sora app and web experience were discontinued on 26 April 2026.
The Sora API was discontinued on 24 September 2026.
Users could export their videos from sora.chatgpt.com/sunset; after any final export window, OpenAI says it will delete the remaining data.
Purchased ChatGPT or Sora credits can still be used for Codex.

That left consumers who used Sora to choose other tools, including xAI’s tool, Google’s Veo, Runway and Pika.

What Grok Imagine can do now

xAI’s generator launched in August 2025 for paying subscribers, turning prompts or still images into short clips with sound. At the time Musk called it “AI Vine”, after the six-second video app Twitter bought in 2012 and closed in 2016 [5]. The name signalled the aim: fast, shareable clips rather than film production.

On 16 June 2026 xAI made its Video 1.5 model generally available on its API and rolled out a faster version in the Grok apps and on grok.com/imagine [2]. The company says Video 1.5 generates sound effects, ambience and dialogue in the same pass, with better motion and physics. It says the Fast version makes a 6-second 720p clip in about 25 seconds, down from more than 40 seconds before [2]. The release also added Projects, parallel generations and library search.

xAI’s developer documentation lists the current options [3]:

Clips from 1 to 15 seconds, from text, a still image or reference images.
Native audio with lip-synced speech on the 1.5 models.
Native 1080p on Video 1.5; the Lite model reaches 1080p by upscaling 720p.
Up to 14 reference images and three preset voices on the full 1.5 model.
Vertical, square and widescreen aspect ratios, including 9:16 for phones.

The Lite model, added in October, is priced from $0.02 per second at 480p, as TSN reported in its AI creative tools round-up.

Grok deepfakes and the safety question

Cheap, fast AI video helps small creators, teachers and charities make explainers, product clips and animations that once needed a studio. The same speed makes abuse easier. Grok’s image tools have faced heavy criticism over sexualised deepfakes of real people, and Ofcom opened an investigation into X on 12 January 2026. In October its chief executive told MPs the legal bar for an urgent court order had been too high, as TSN reported. That investigation has made no finding yet.

Grok Imagine and Sora: speed versus control in AI video

OpenAI’s exit and xAI’s releases show two different bets on consumer AI video. OpenAI ended a standalone video app and kept its focus on ChatGPT, Codex and enterprise products. xAI went the other way: it put video generation inside Grok and X, made clips faster and cheaper, and opened the same models to developers through an API.

For everyday creators, that is mostly good news. A 15-second clip with matching sound can now be made on a phone in under a minute, at a price per second that small businesses can plan around. Developers can use the same models in their own apps instead of relying on one social feed. Competition from Google, Runway, Kling and Vidu also keeps prices falling, which Runway’s move to host Kling models shows.

The costs are real too. Clips made inside X are tied to that platform’s rules and moderation. Watermarking and provenance labels matter more as realistic clips get cheaper, and regulators are still testing whether their powers work at internet speed. Musk’s promised “big” update did arrive as Video 1.5, but claims that xAI would beat rivals on quality are still marketing; independent, like-for-like comparisons remain scarce.

The useful test for readers is simple: pick the tool whose terms, export options and safety controls you can live with, and keep your own copies of anything you make. Former users of OpenAI’s app learned that a product can close within weeks.