Cicadas.ai
Back to blog

Model updates

Grok Imagine Odyssey: What the AI Odyssey Series Is, and How to Make a Homeric Scene

What is known about xAI's Grok Imagine adaptation of Homer's Odyssey (pilot teaser, episode 1, the feature-length promise), where to watch it, which model it reportedly uses, and how to make your own short Homeric scene on Grok Imagine Video 1.5.

Oct 28, 2026By Kevin Fan

Video examples here were generated on Cicadas and keep their model, settings and date. Cicadas earns money when you generate with these models. How we test

Test window:
2026-09-27 – 2026-10-02
Clips recorded with these models:
7
Models:
Grok Imagine Video 1.5, Grok Imagine Video 1.5 Lite
Their provider cost:
≈$2.61
Selection:
every clip we keep is listed with its settings, cost and request ID; failed or discarded runs are listed in the test record.
Grok Imagine Odyssey: What the AI Odyssey Series Is, and How to Make a Homeric Scene

Short answer: "Grok Imagine Odyssey" is the name people are searching after xAI's Grok Imagine began releasing an AI-generated adaptation of Homer's Odyssey: a pilot teaser on September 29, 2026, episode 1 on October 23, 2026, and a promise from Elon Musk of a feature-length version before the end of the year. The reports say the pilot was made with Grok Imagine Video 1.5. Cicadas runs that model (and its Lite tier) through fal.ai on the Grok Imagine page, 5 or 10 seconds at a time with sound, from 25 credits a clip, which is enough for a ship on a wine-dark sea or a loom in a quiet hall, but not for an episode.

We make Cicadas. We are not xAI, we had no part in the Odyssey series, and we cannot show it: there is no footage, no still and no image of anyone from it on this page. What we can do is say what has been reported about the project, with dates and links, say exactly which Grok Imagine model this site runs and what it costs, and show what our own recorded Grok clips do and do not prove about scenes like these. We have not watched the episode ourselves and we write the reported facts as reported.

1. What the Grok Imagine Odyssey project is

Here is the timeline as the reports describe it. We read these reports on October 28, 2026 as search summaries, not the full pages, and we did not open the X posts they refer to, so each line names its source.

What we could not confirm, and so do not say: the episode's running time, who voices it, how it was scored, whether the September teaser and the "trailer" are the same clip, and whether episode 1 was made with the same model build as the pilot. If you want the primary source, the reports point to the Grok account on X.

Where to watch it. According to the three reports above, on X, through the Grok account's post and Musk's repost of it. We link the reports and not the posts because the reports are what we read. The series is xAI's work; nothing on Cicadas reproduces it.

2. Which Grok Imagine model the series reportedly uses, and what Cicadas runs

The Basenor report names Grok Imagine Video 1.5 for the pilot. That is a model xAI announced on June 16, 2026, and it is one of the two Grok tiers on this site. Cicadas is an independent site, not affiliated with or endorsed by xAI; it calls the model through fal.ai, so you need a Cicadas account and credits, not an X or Grok account.

| On Cicadas | Endpoint (fal.ai) | Length | Resolution | Sound | 5 s from text | |---|---|---|---|---|---| | Grok Imagine Video 1.5 | xai/grok-imagine-video/v1.5/text-to-video and /image-to-video | 5 or 10 s | 480p or 720p | always on | 100 credits at 480p, 175 at 720p | | Grok Imagine Video 1.5 Lite | xai/grok-imagine-video/v1.5/lite/text-to-video and /image-to-video | 5 or 10 s | 480p or 720p | always on | 25 credits at 480p, 38 at 720p |

Both run on the Grok Imagine page. Text mode frames 16:9, 9:16 or 1:1; image mode keeps the shape of the picture you upload and adds 3 credits to the quote. Prompts take up to 2,000 characters. The first credit pack is 1,000 credits for $9.90, so 100 credits is about $0.99, 175 about $1.73, and 10 seconds at 720p on the full model (350 credits) about $3.47.

Two honest gaps. We do not know which settings, how many takes, what editing or what voice work went into the series; a 5-second clip from a form is not the same process as an episode. And "Grok Imagine Video 1.5" on grok.com, in the Grok apps and through xAI's API is xAI's own product with xAI's own limits, which we have not checked; this page is about the fal.ai version Cicadas runs.

3. What our recorded Grok clips prove, and what they do not

Every Grok clip on this site was generated on the Grok Imagine page between September 27 and October 2, 2026, each with its prompt, settings and request ID kept. None of them is a Homeric scene: there are no ships, crowds, armour, dialogue or period costume among them, and we have not made an Odyssey-style clip for this post, because we spend on a clip only when a page needs it. So the clips below are evidence about the model's habits, not about epics.

The closest thing we have to a lone traveller crossing a landscape is this text-to-video clip at 720p, 5 seconds, generated on September 27, 2026:

The prompt asked for an astronaut walking across a rust-red dune at dawn, dust lifting with each step, a side tracking shot, "wind and footsteps crunching in the sand", one continuous shot. The walk and the dust trail hold together for the whole clip. The audio, though, is calm music: the wind and footsteps did not arrive. That is the pattern across our clips. The bamboo-forest tea clip (September 30) did get the birdsong it asked for; the Lite kettle clip (October 2) asked for a whistle and got a very quiet steady noise; the three Lite clips are all very quiet.

The second habit matters more for a story scene. This clip, 720p, 5 seconds, September 29, 2026, asked for one continuous shot:

Grok cut between three camera angles instead. For a product shot that is a fault; for a Homeric scene it may be a gift, since a cut from the ship to the rowers to the horizon reads like editing. You cannot ask for it reliably, and you cannot forbid it reliably either.

So: a single subject in a landscape with a slow camera move holds; a stylised look (the rooftop robot, 480p, September 29) holds; sound cues are unreliable; "one shot" is a request, not a setting. What the clips do not tell you is how the model handles several people, a crowd of rowers, a giant, or anyone speaking. Treat the next section as a prompt shape with that unknown in it.

4. Making your own short Homeric scene

Homer's Odyssey is in the public domain, so the sea, the cyclops, the suitors and the loom are yours to use. What is not yours: characters, costumes or shots from any film or series, including the Grok one; the likeness of any actor or real person (the site refuses prompts that name real people); and any song. Make a scene from the book, not from someone's adaptation of it.

The shape that matched our recorded clips best: one subject, one action, one camera move, the look, then a sound cue you are willing to lose, in one paragraph, with "one continuous shot, no text, no logos" at the end. Three starting points, untested on this site:

  • The wine-dark sea. "A single ancient Greek galley with one square sail and twenty oars rows across a dark purple sea at dusk, white foam at the bow, a rocky island far ahead. Slow aerial drone shot circling the ship, golden rim light, painterly realism. Oars creaking and waves. One continuous shot, no text, no logos."
  • The cyclops cave. "Inside a vast firelit cave, a fictional one-eyed giant sleeps against the wall while a small group of sailors in rough wool tunics creep past sheep toward the open mouth of the cave. Low slow dolly along the sailors, flickering fire light, deep shadows. Crackling fire and distant breathing. One continuous shot, no text, no logos."
  • Penelope's loom. "A fictional woman in a plain linen dress weaves at a tall wooden loom in a stone hall at night, oil lamps on the walls, a half-finished shroud on the frame. Slow push-in from the doorway to her hands, warm lamp light, quiet. The clack of the loom and wind outside. One continuous shot, no text, no logos."

Settings we would start with: text mode, 16:9, 5 seconds, 480p on Grok Imagine Video 1.5 Lite (25 credits, about $0.25) to check whether the scene reads at all, then the same prompt on Grok Imagine Video 1.5 at 720p (175 credits, about $1.73) once it does. If you have a painting or drawing you own of a ship or a hall, image mode will animate it and keep its shape; it adds 3 credits. For a quick first draft of a Grok prompt in this order, the free Grok Imagine prompts builder writes it in your browser.

Limits to know before you spend:

  • 5 or 10 seconds, then another clip. There is no episode mode; a sequence is several clips joined in your editor. Seedance 2.5 makes single clips up to 30 seconds if one long take matters more than the Grok look.
  • Sound is always on and not reliable. Expect music where you asked for oars; add your own sound in the edit.
  • No dialogue promise. We have not recorded a speaking Grok clip, so we cannot tell you how lines come out.
  • No names, no faces. Write "a fictional sailor", not an actor; do not upload anyone's photo without their permission.
  • Text in the frame is unreliable. Put titles on in your editor.
  • Charged when delivered, refunded when it fails. A clip that arrives is charged even if the giant has two eyes; a failed generation or file check returns the credits.

If Grok's look is not what the scene needs, the Grok Imagine alternative page runs the same kind of prompt on MiniMax H3 Max, Seedance 2.5 and Kling 3.0 Pro from one form.

Sources and method

Reports: Basenor, TeslaNorth, October 25, 2026 and Khaleej Times, seen as search summaries on October 28, 2026 (the full pages were not fetched in the session that wrote this), plus a NetflixJunkie report on the bloopers. Model facts, prices and clip details are from the site's own generator settings and recorded clips on October 28, 2026; the working note with each claim, its source and the trend-board row that prompted this post is kept in the site repository at docs/research/claude-blog-20261028/grok-imagine-odyssey.md. The search term itself appeared on the Google Trends board we watch on October 25, 2026, two days after episode 1.

Related