Grok Imagine releases episode 1 of its AI-made Odyssey with Wonder Studios

What xAI released

Grok Imagine, the image and video generator from xAI, has released the first episode of an AI-made series adapting Homer's Odyssey. The Grok Imagine account posted it on X on 23 October 2026 with the line "Watch Episode 1 of our Odyssey series made with Grok Imagine," and credited a creative collaboration with the London studio Wonder Studios and director Dev Walker. Wonder Studios reposted it with its own tagline: "Homer didn't have Grok... we do."

Elon Musk shared the episode the same day. Mashable's review says it runs a little over seven minutes and covers the Greeks' arrival at Troy, and Magica's report puts the running time at 7 minutes 12 seconds, citing release coverage. The episode is free to watch on X, and the post asks viewers to say in the comments whether they want a full breakdown of how it was made.

The release is labelled Episode 1, which makes it the opening installment of a series rather than a single film. Neither xAI nor Wonder has said how many episodes are planned, how long the finished adaptation will be, or when the next part will appear.

Where the project came from

The series follows a public pledge. According to Mashable, Musk posted on X in July that Grok Imagine would make a full-length Odyssey movie before the end of the year that would be "historically accurate and true to the art of Homer." That makes this episode the first concrete step toward a deadline that is about two months away.

The promise grew out of Musk's months of criticism of Christopher Nolan's film adaptation of The Odyssey, in particular its casting. Mashable notes that Nolan's film opened to $124.5 million in July, has grossed more than $1.7 billion worldwide and is due for digital release on 17 November. Magica points out that a box-office total for a studio feature is not a like-for-like comparison with a free online episode, and the same caution applies to any claim about which version is more faithful to a poem built on myth.

Nolan has been openly sceptical of generative AI. Mashable cites a Guardian report of him telling AFP that the idea of AI replacing human creativity is "a nonsense." The Grok project therefore doubles as a rebuttal and a product demonstration, with Musk using a rival adaptation as the backdrop for showing what xAI's video model can do.

Who made it

Wonder Studios is not a newcomer assembling clips for the first time. TechCrunch reported in October 2025 that the London-based studio raised a $12 million seed round led by Atomico, with LocalGlobe and Blackbird, after a pre-seed round that included executives from ElevenLabs, Google DeepMind and OpenAI. It had made an AI music video for Lewis Capaldi with DeepMind, YouTube and Universal Music Group, and released an anthology series called Beyond the Loop.

Magica, summarising Wonder's own account of its first year, says the studio worked with fal, Freepik, Kling and ElevenLabs on short films and produced an animated series based on the Nordic children's books Maxi and Helium. Wonder describes itself as agnostic about which models it uses and has said it aims to co-own original intellectual property with the creators it backs, though those general terms do not tell us who owns Odyssey.

Dev Walker, the credited director, is a generative AI filmmaker. Mashable says both Walker and Wonder specialise in that kind of work. In practice the episode is a joint showcase: xAI supplies the model and the distribution on X, and Wonder supplies a production team that already knows how to turn thousands of generated shots into something with an edit, a soundtrack and a running order.

The tools behind it

The announcement does not say which version of Grok Imagine produced the episode. The most recent public model is Grok Imagine Video 1.5, which xAI made generally available on 16 June. xAI's post says it improves motion, physics and audio, with sound effects, ambience and dialogue generated in the same pass as the picture. Video 1.5 Fast makes a six-second 720p clip in about 25 seconds, down from more than 40 seconds for the previous model.

The same June post featured an earlier Odyssey: a trailer that xAI says David Thompson made entirely with Grok Imagine 1.5, with a behind-the-scenes video. xAI also added Projects for organising work, multiple agents that run several prompts at once, and search across a user's library of generated images and videos. Developers reach the model through the xAI API as grok-imagine-video-1.5, which takes a starting image and a description of the motion.

Those features hint at how a seven-minute piece gets made. Each generated clip is a few seconds long, so a long episode means generating many candidates, choosing the usable ones and cutting them together. Magica cites a production account in which Wonder's creators said an earlier Odyssey pilot took 15 days and drew on thousands of generated images and clips. That figure describes the pilot, not the new episode, and neither company has published a budget.

What reviewers saw

Early reviews are mixed at best. Mashable says the episode has the familiar problems of generated video: unnaturally perfect teeth, hair that looks weightless and almost liquid, smudgy skin that suggests the model does not really understand light, and backgrounds that resemble the texture pop-in of a video game loading detail late.

The storytelling drew more criticism than the pictures. The reviewer found the characters hard to tell apart, some of them looking very like real people, and noted that Odysseus is the only character named in the whole episode. Mashable also says every character is white. Magica adds that viewer reactions included praise for atmosphere and visual consistency alongside complaints about rapid cuts, facial expressions and voices.

These are qualitative judgements of one episode, not a controlled comparison with other video models. They do line up with the weaknesses most AI video still shows once clips are strung together: keeping faces and costumes consistent across shots, and giving a story enough structure that a viewer can follow who is who.

What the episode does not answer

The biggest gap is cost. Generative video is often sold on the promise of making film far cheaper, but the release gives no figures for generation time, compute, discarded material or the human hours spent on writing, editing, sound and direction. Magica notes that Wonder has said its key internal measure is cost per minute, combining API charges with human workflow time, yet no such number has been shared for Odyssey.

Nor is it clear how the series will meet Musk's pledge. A year-end deadline and a seven-minute opening episode leave a long way to go if the goal is a feature-length film, unless episodes are released much faster or later compiled into one. The breakdown that Grok Imagine offered in its post could answer some of these questions if it is published.

Why it matters for AI video

The Odyssey episode lands in a crowded month for AI video. Kuaishou made Kling 4.0 fully available with 30-second clips, HeyGen launched a model that makes picture and sound in one pass, and the creator platform Higgsfield passed $1 billion in annualised revenue. Most of that activity is about short clips for ads and social posts. A narrative series with a named director is a different test, and a public one.

It also raises questions about labelling and trust. The episode is openly presented as AI-made, but the same tools can produce footage that is not, which is why states are starting to regulate it. Pennsylvania, for example, now requires advertisers to label AI-generated images, video and audio. Characters that look like real people, as Mashable describes, are exactly the kind of output that makes those rules matter.

What to watch next

The next milestones are simple to track: a second episode, a release schedule and the promised production breakdown. If xAI and Wonder publish the tools, the number of generations and the human work behind each finished minute, the project becomes a useful case study in what AI filmmaking costs today. Without them, it remains a well-promoted demonstration.

The creative test is continuity. Viewers will judge whether later episodes keep characters recognisable, name them clearly and carry a story across many scenes, which is where generated video has struggled most. That, more than the speed of any single clip, will decide whether a full AI-made Odyssey arrives by the end of the year in a form people want to watch.