Grok Imagine Video 1.5 Lite hits the API: workhorse text/image-to-video from $0.02/s at 480p, about a quarter of Video 1.5's list price

Core Background & Industry Pain Points

Video APIs bill per second, so flagship models get expensive fast for bulk shorts, ad variants, or product demos, and many simple shots need no references, voices, or keyframes. On 2026-10-08 Grok Imagine launched Video 1.5 Lite in the Grok Imagine API as the workhorse for text-to-video and image-to-video (model docs).

Architecture Highlights & Internals

grok-imagine-video-1.5-lite takes text or an image and outputs video with native audio and lip-synced speech, 1–15 s, at 480p/720p/1080p (1080p is upscaled from 720p, not native), in 9 aspect ratios, with image-to-video matching the input's ratio. Unlike 1.5, Lite has no reference images (1.5 takes up to 14), voice references, first/last frames, or keyframes. xAI recommends Lite for simple videos and 1.5 for references, voices, pinned frames, or native 1080p (capability table). Batch API supported; async submit, poll by request_id, download; 10 RPS; us-east-1 and us-west-2.

Authoritative Benchmarks & Measured Scores

No quality evals or blind Elo were published, only pricing: $0.02/s at 480p, $0.03/s at 720p, $0.14/s at 1080p. On model-page list prices, Lite is $0.020/s vs Video 1.5 at $0.080/s, about a quarter. A 10 s 720p clip costs about $0.30. No quality scores invented.

Developer Hands-on Guide

1. Set model to grok-imagine-video-1.5-lite; the xAI SDK and Vercel AI SDK poll for you. 2. Draft at 480p/720p, then go to 1080p or 1.5 for finals. 3. For character consistency or frame pinning, use 1.5, since Lite does not support them. 4. Use the Batch API for volume and respect the 10 RPS limit.