What’s new
- Pure text-to-video, no starting image required: at launch in June, 1.5 could only animate a starting image plus a prompt; the July 31 update added text-to-video, so a written description alone is now enough.
- Native 1080p: the June preview and the general-availability release were both capped at 720p; the July 31 update brought native 1080p to both text-to-video and image-to-video.
- Image and voice references together, up to seven per generation: a reference image can lock in a face, a product, or a location, and a voice reference keeps the voice consistent throughout the clip.
- Faster generation with steadier motion and physics: xAI says a 6-second, 720p clip now renders in about 25 seconds, down from 40-plus seconds previously, with better motion, physics, and audio.
What it’s good for
- Clips where a character or product has to look the same throughout: feed it reference images and a voice reference together; xAI lists virtual try-on, product placement, and character-consistent storytelling as use cases.
- A quick concept clip when you don't have a starting image: describe the shot in a prompt and it generates directly from text, no reference image needed.
- Touching up or continuing a clip you already have: use video editing to change one detail (add an accessory, change a color) or video extension to keep the action going from the last frame, instead of regenerating the whole thing.
Prompting tips
- Describe the camera move, the pacing, and the sound design: that's xAI's own framing for what a prompt should cover.
- Write one flowing, cinematic scene description instead of a list of keywords: xAI's own examples read like a single sentence tying subject, action, setting, and mood together — for example, "a glowing crystal-powered rocket launching from the red dunes of Mars, ancient alien ruins lighting up in the background."
- Tag reference images and voices by position: use
<IMAGE_0>,<IMAGE_1>,<AUDIO_0>in the prompt to point at which uploaded reference you mean. - State the aspect ratio if you want one other than your source image's: image-to-video defaults to the input image's aspect ratio unless you set it explicitly.
- Sound is on by default, so write what you want to hear: effects, ambience, and dialogue are generated in the same pass as the picture — describe them ("rain," "a door latch," a line of dialogue) rather than leaving them out, or turn audio off if you don't want it.
- Keep prompts only as complex as they need to be: xAI's docs note that prompt complexity affects processing time.
Vertical 9:16. The girl in <IMAGE_0> keeps the same face and the same red hoodie throughout.
An alley at dawn, lanterns strung overhead. She jogs down the narrow street, the camera tracking her in a slow lateral pan, and glances back with a smile as she turns the corner.
Sound: footsteps, distant firecrackers, the soft rustle of lanterns swaying. No face changes, no extra people.
Known limits
- Editing and extension inherit the source clip's limits: video editing and video extension keep the original video's duration and aspect ratio and cap out at 720p, even though a fresh generation can reach 1080p.
- You can't clone your own voice yet: voice references are drawn from a small set of preset voices (up to three per generation); using your own recorded audio as a voice reference is, per xAI's docs, available to trusted partners on request.
- Reference timing is coarse: reference-to-video generation allows at most four keyframes per request, and timestamps snap to a 1/3-second grid.
Using it in Nomi
On the generation canvas, add a video card, pick “Grok Imagine 1.5” in the model picker and choose a mode (Text to video, First frame to video). Add your references, write the prompt and generate. In the storyboard you can pick it per shot. The first time, connect any one of the providers above under Settings → Models.
Sources
- Grok Imagine 1.5 Preview (xAI news, 2026-06-03)
- Grok Imagine Video 1.5 (xAI news, general availability, 2026-06-16)
- Imagine Video 1.5 with References (xAI news, 2026-07-31)
- Video Generation (xAI developer docs)
- Reference-to-Video (xAI developer docs)
- Video Editing (xAI developer docs)
- Video Extension (xAI developer docs)
- Release Notes (xAI developer docs, July 2026 entry)
