xAI has introduced Grok Imagine Video 1.5 Lite, a distilled variant of its video model designed for faster generation. Users can create short clips from text prompts or images, complete with native audio. It joins the full Grok Imagine Video 1.5, although Artificial Analysis rates it below that version in most of the capabilities assessed.
Clips up to 15 seconds and seven aspect ratios
Lite can generate video from either a text prompt or an image. Users can set the clip duration anywhere from 1 to 15 seconds, with native audio also supported.
Resolution options are 480p, 720p and 1080p. The model offers seven aspect ratios, including the widescreen 16:9 format and vertical 9:16.
According to the published model description, video labeled as 1080p is initially rendered at 720p before being upscaled. Its final frames have 1080p dimensions, while the native image detail originates at 720p.
Artificial Analysis ranks Lite above Veo 3.1, but below the full model in capabilities
At the time of publication, Artificial Analysis placed Lite 17th in both its AA-Video-T2V v2.0 and AA-Video-T2V-Silent v2.0 rankings. That put it two places ahead of Google’s Veo 3.1 in the reported comparison.
For a ten-second clip at 1080p, Artificial Analysis recorded a median generation time of 60.5 seconds. Its speed comparison covered 12 models and placed Lite on the quality-speed Pareto frontier, meaning none of the compared models surpassed it in both quality and speed.
The comparison against the full Grok Imagine Video 1.5 reveals the lightweight version’s trade-off. Artificial Analysis found equal scores in Physics, but lower scores for Lite in the other nine capability groups. Multi-Scene & Narrative showed the smallest gap between the versions.
Artificial Analysis finds stronger scene continuity and lighting, weaker dialogue
The AA-Video-T2V v2.0 benchmark assesses ten capabilities and ten types of use. Artificial Analysis found Lite closest to the quality frontier in Multi-Scene & Narrative, Lighting & Materials and Text Rendering. These categories examine aspects including continuity between shots, lighting and how text is displayed.
The evaluator identified the largest shortcomings in the categories covering dialogue and lip synchronization, Dialogue & Lip Sync, and human anatomy, Human Anatomy.
Artificial Analysis’s use-case results were most favorable for Architecture & Real Estate, Consumer content and Productivity & Knowledge Work. These groups cover interior walkthroughs and virtual staging, alongside personal videos and supplementary footage. Explainer clips and animated data visualizations are also included in the assessment.
Access through xAI API, fal and Vercel AI Gateway
Developers can access the model through the xAI API, fal or Vercel AI Gateway. Vercel makes it available using an experimental video helper in its AI SDK.
xAI charges $0.14 per second of 1080p output, $0.03 at 720p and $0.02 at 480p; the 1080p rate is 44% below that of the full Grok Imagine Video 1.5 and 65% below that of Veo 3.1 with audio.
Available settings depend on the provider. Request fields may differ between xAI, fal and Vercel, while some third-party catalogs list only 480p and 720p options for the Lite model.
