AI movie models: xAI Grok Imagine

aitamer.news is produced with Grok Bot, an xAI product. xAI has no say over this review.

AI movie models, episode 3. Previously: Kling and Gemini Omni vs Utopai X.

This series explains one or two AI video models per post, using the same twelve questions every time. Facts come from the vendor’s own pages unless marked as a user report or our opinion. Prices were checked on 28 October 2026.

Grok Imagine is the image and video generator from xAI, whose developer docs and terms now carry the SpaceXSI name. The current video model is Grok Imagine Video 1.5 (grok-imagine-video-1.5): xAI announced a preview on 11 June 2026, the full model on 17 June and reference support on 7 August. A cheaper 1.5 Lite followed (our Lite story), and the older Grok Imagine Video 1.0 still handles editing and extension.

What it does differently. Breadth of control in one call: up to 14 reference images, up to 3 preset voices, a first and last frame plus up to 4 keyframes, nine aspect ratios and native audio with lip sync (xAI docs). It is also built into the Grok and X apps, so many creators reach it through a subscription they already have.

Grok Imagine Video 1.5 against the checklist

1. Resolution. 480p, 720p and native 1080p for 1.5. Lite outputs 1080p upscaled from 720p. Video 1.0 goes up to 720p. Clips made with references are capped at 720p (xAI docs).

2. Aspect ratios. 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9 and 5:2 for 1.5.

3. Reference types. Reference images (people, objects, clothing; they guide the clip without locking the first frame), preset voice references, first frame, last frame and keyframes. Video-to-video editing and extension run on Video 1.0. Lite does text-to-video and image-to-video with no references.

4. References at once. Up to 14 reference images and up to 3 preset voices; a reference-driven clip runs up to 15 seconds at up to 720p. First and last frame plus up to 4 keyframes. Video 1.0 takes up to 7 references for editing.

5. Price (checked 28 October). xAI’s 1.5 model page lists $0.08, $0.14 and $0.25 a second at 480p, 720p and 1080p, plus $0.01 per input image; preset voice audio is free. Lite: $0.02, $0.03 and $0.14 a second. From those numbers, a 5-second 720p clip costs $0.70 and a 10-second 1080p clip $2.50 on 1.5 (our arithmetic). Video 1.0 costs $0.05 a second at 480p and $0.07 at 720p, plus $0.01 per second of input video. xAI’s pricing index now lists the same per-resolution rates; the $0.08 headline figure is the 480p rate. Consumer plans: Free, Plus $8 a month (includes Imagine), Premium $30, Super $100 and Ultra $200, with usage pooled across features; video allowances per plan are not published.

6. Performance. xAI does not publish generation times. On the Artificial Analysis text-to-video leaderboard (now AA-Video-T2V v2.1, read 28 October), 1.5 is #12 (Elo 1057) and Lite #19 (990), with Wan 3.0 first (1174). Reseller Runware says Lite makes a 5-second image-to-video clip in about 16 to 32 seconds (post); that is its figure, not xAI’s.

7. Character consistency. Reference images carry a person, outfit or product into new clips, and preset voices keep how they sound; xAI suggests this for “stories with recurring characters”.

8. Scene consistency. Clips on 1.5 run 1 to 15 seconds. Video 1.0 extends a clip by 2 to 10 seconds per step and edits a source clip of up to 8.7 seconds. First, last and keyframes pin a scene’s layout. No multi-shot mode is documented.

9. Compared with the previous model. 1.5 adds native 1080p, references with voices, keyframes and more aspect ratios over 1.0, which tops out at 720p. On Artificial Analysis it ranks well above Lite; 1.0 is kept for editing and extension.

10. What users think (user reports, not facts).

• Praise: @mustansant made a short film from his X Premium subscription and says a 6-second 720p clip rendered “in about 10 seconds”. @HighworthAI ranked Grok “a close #2” behind Wan 3.0 in a Hindi-prompt test. @kenn_ronin says it is “getting better”.
• Complaints: @mathieuhq calls the headline Lite price “un peu trompeur” (a little misleading), since $0.02 a second is the 480p rate and 1080p costs $0.14.

11. General quality. We have not generated Grok Imagine clips for this series. Its blind-vote rank sits between Gemini Omni and Kling 3.0. xAI’s Grok Imagine page and its 1.5 announcement carry official examples.

12. Our verdict (opinion). Pick Grok Imagine 1.5 if you want many references, unusual aspect ratios such as 21:9 or 5:2, and an API with per-second prices, or if you already pay for X or Grok. Skip it for long takes (15 seconds maximum) or if top blind-vote quality matters most: on Artificial Analysis, Wan 3.0, Utopai X and Gemini Omni rank higher.

Rights and licensing

Under xAI’s consumer terms (last updated 27 October 2026), the heading still reads “You Own Your User Content”, covering input and output. But the 27 October text now says that “When using Output or SpaceXSI’s name, logos, trademarks, or other brand elements, you are required to obtain our permission and attribute your generation of the Output to the Service, as detailed in our Brand Guidelines”. Read literally, that asks for permission as well as attribution, so check the Brand Guidelines before commercial use. A new clause also says you may be required to apply an AI disclosure to output, or SpaceXSI may apply one. Images and videos from the Grok apps carry a Grok watermark; xAI’s FAQ says there is no setting to remove it and removing it is prohibited under the Acceptable Use Policy. API use falls under xAI’s enterprise terms. No open weights for the video models.

Summary

#	Checklist	Grok Imagine Video 1.5	1.5 Lite
1	Resolution	480p, 720p, native 1080p (720p with references)	1080p upscaled from 720p
2	Aspect ratios	9, from 21:9 and 5:2 to 9:16	not separately listed
3	Reference types	images, preset voices, first/last frame, keyframes	none
4	References at once	14 images, 3 voices, 4 keyframes	none
5	Price (28 Oct)	$0.08 to $0.25/s (model page); $0.70 per 5 s at 720p	$0.02 to $0.14/s
6	Performance	time not published; AA #12 (1057)	AA #19 (990)
7	Character consistency	references and voices	not supported
8	Scene consistency	1 to 15 s; extend via 1.0	length not on model page
9	vs 1.0	1080p, references, keyframes	cheaper, faster distill
10	Users	fast renders praised; price headline criticised	pricing confusion
11	Quality	not tested by us	not tested by us
12	Verdict (opinion)	control and formats	cheap drafts