Grok Imagine Video Pricing: What AI Video Really Costs Per Second in 2026

xAI's standard Grok Imagine Video API charges $0.05 per second at 480p and $0.07 per second at 720p. The newer Imagine Video 1.5 tier charges more — $0.08/sec at 480p, $0.14/sec at 720p, and $0.25/sec at 1080p — for higher output quality. Both models cap a single generation at 15 seconds.

Those numbers describe the cost of a generated second. They don't describe the cost of a second you actually keep. That gap — between generated cost and accepted cost — is the real number this guide works out.

The full pricing table

FACT: Because a single generation tops out at 15 seconds, any finished sequence longer than that requires multiple separate generated clips stitched together — there's no single-call option for a 30- or 60-second continuous video.

Basic cost per clip, before regeneration

CALCULATION for a 30-second finished sequence, built from two 15-second clips at Imagine 1.5, 720p, first-try success: 2 × $2.10 = $4.20 raw generation cost. For a 60-second finished video, four 15-second clips: 4 × $2.10 = $8.40.

These are floor numbers — the cost if every single generation is accepted on the first attempt. In practice, it isn't.

Raw cost per generated second vs. effective cost per accepted second

This is the calculation that actually matters for budgeting a real project. A generation that gets rejected — wrong motion, an artifact, a prompt that didn't land — still cost money to produce. The effective cost per accepted second is the raw per-second price multiplied by how many attempts it typically takes to get one you'll actually use.

CALCULATION: at a 3x regeneration rate on Imagine 1.5 720p, a 15-second accepted clip actually cost 3 × $2.10 = $6.30 in generation spend to produce one $2.10-labeled clip — the other two generations were discarded, but xAI billed for all three, since a generated-but-rejected video is still billed under standard usage policy (a request that fails validation before any video is produced is not billed; a video that renders but gets discarded by you or blocked by content policy afterward still is).

Real project examples: 30 and 60 seconds of finished footage

30-second finished sequence (two 15-second clips), Imagine 1.5, 720p

60-second finished video (four 15-second clips), Imagine 1.5, 720p

CALCULATION check (60-second, 3x row): 4 clips × 3 attempts each = 12 generations × 15 sec × $0.14/sec = 12 × $2.10 = $25.20. At a realistic mid-range regeneration rate (many teams report needing several attempts per usable clip on current-generation AI video tools), a 60-second finished piece at 720p on Imagine 1.5 lands somewhere in the $17–25 range — a long way from the $8.40 "ideal" number most pricing pages implicitly promise.

Resolution's compounding effect on regeneration cost

Regeneration rate and resolution multiply against each other, not add. A 1080p clip costs more per second to generate and, if it's more prone to needing regeneration at that quality bar for a given prompt complexity, compounds further.

CALCULATION — one 15-second clip at 1080p ($0.25/sec) at increasing regeneration rates: 1x = $3.75, 2x = $7.50, 3x = $11.25, 5x = $18.75. Compare that 5x-row figure to the same clip at 480p standard tier, 5x regeneration: 15 × $0.05 × 5 = $3.75 — exactly what the 1080p clip costs at a single successful attempt. Resolution choice and regeneration rate aren't independent decisions; a lower-resolution tier with a higher regeneration tolerance can land at the same real cost as a premium tier that needs fewer attempts, or vice versa, depending on your actual results.

Practical ways to lower effective cost per accepted second

• Iterate on the prompt at the cheapest tier first. Since 480p standard is roughly a third of the 1080p 1.5 rate, refining a prompt's motion and composition at the low tier before committing to a higher-resolution final pass avoids paying premium rates for early, exploratory attempts.
• Batch similar shots together. If a project needs several clips with a consistent style or character, working through prompt variations in one focused session — rather than starting cold on each — tends to reduce the discard rate as the working prompt pattern is refined.
• Separate "must be perfect" shots from "good enough" shots within the same project, and apply a higher resolution/regeneration budget only to the former. Not every second of a finished video carries the same acceptance bar, and treating them identically overspends on the low-stakes ones.
• Track your own regeneration rate from the first project onward. Because no vendor publishes this number, it's the single most valuable piece of cost data a team can generate for itself — and it turns every subsequent budget estimate from a guess into a measurement.

Comparison context: where Grok Imagine Video sits

For a provider-specific comparison of another low-cost video model, see the Creatify Boreal AI video cost guide and compare usable-clip assumptions, not just the headline per-second rate.

Providing exact competitor pricing without an equivalent billing model risks a false apples-to-apples comparison — some competing video APIs bill per generation regardless of length, others per second, others on a credit system with opaque conversion rates. What can be said reliably: Grok Imagine Video's standard tier, at $0.05–0.07/sec, sits at the low end of currently available API-accessible video generation pricing, and the 1.5 tier at $0.08–0.25/sec (depending on resolution) sits in the middle of the market — priced above the cheapest available options but below premium cinematic-grade video models that can run several times higher per second. Where exactly a specific competitor lands depends on its own billing structure and resolution/duration limits, which should be checked directly rather than assumed from a single blended number.

Who this makes financial sense for

Creators and small teams producing high volumes of short social clips, product teaser videos, or ad variants, where the 480p or 720p standard tier's low per-second cost tolerates a higher regeneration rate without the total spend becoming prohibitive — and where the 15-second-per-clip ceiling fits the actual content format (most social video is well under 15 seconds per shot anyway).

Who should choose alternatives

Projects needing native 1080p from the standard tier (not offered — 1080p is only available on the pricier 1.5 tier), continuous single-shot footage longer than 15 seconds without visible stitching seams, or projects where regeneration rate on this specific tool has proven high enough in testing that the effective cost per accepted second no longer beats an alternative provider or a traditional production approach for the same output quality bar.

FAQ

What's the real cost of a finished 60-second video? At the "ideal" first-try rate on Imagine 1.5 720p, $8.40. At a realistic 2–3x regeneration rate, more like $17–25. The gap between those numbers is the actual cost of the tool, not the number on the pricing page.

Does xAI charge for rejected or unusable generations? Yes, for anything that actually renders. A request that fails validation before any video is produced isn't billed, but a video that generates successfully and is then discarded by you, or blocked by content moderation after generating, is still billed under standard usage terms.

Why doesn't the standard tier offer 1080p? xAI reserves 1080p output for the newer, pricier Imagine Video 1.5 tier; the standard tier tops out at 720p.

Is a higher resolution always a worse deal once you factor in regeneration? Not necessarily — it depends on whether the higher-resolution tier actually needs fewer regenerations to hit an acceptable result for your use case. This guide shows the effective-cost math; it doesn't assume a universal answer, because that answer depends on your own prompts and acceptance bar.

How do I make one continuous video longer than 15 seconds? You generate multiple separate clips (each capped at 15 seconds) and stitch them together in a separate editing step — there's no single API call that produces continuous footage past that limit.

Does stitching multiple clips together cost extra on top of generation? Not through xAI's API directly, but yes in practice — assembling separate clips into one continuous video requires editing software or a video-processing pipeline, which is either your own time, a paid editing tool subscription, or a freelance editing cost, none of which appear in the per-second generation price.

Is the standard tier or the 1.5 tier the better default choice? Neither is universally better — it depends on whether the 1.5 tier's higher quality reduces your regeneration rate enough to offset its higher per-second price. If a prompt type reliably needs fewer attempts on 1.5 to reach an acceptable result, it can end up cheaper per accepted second than the "cheaper" standard tier despite the higher sticker price — the same principle that applies when comparing different AI models on a cost-per-successful-task basis rather than cost-per-token alone.

The NotCheapAI verdict

Grok Imagine Video's per-second price is genuinely low, and for high-volume short-form content it can be one of the cheaper API-accessible video options available. But the sticker price describes a generated second, not an accepted one — and the real cost of a finished video depends entirely on a regeneration rate that isn't published anywhere and has to come from your own testing. Before committing a project budget to a per-second number, generate a representative sample at your actual quality bar and count how many attempts it really takes. That number, not the price sheet, is what your invoice will reflect.