Imagine Video 1.5 Multi-Ref and 1080p: Character Consistency Becomes the Product
xAI upgraded Imagine Video 1.5 with text, image, and voice references (up to seven image refs), native 1080p, and prompt-only text-to-video—rolling first to US SuperGrok Heavy and Plus. The trade is production-ready character consistency, not another one-shot clip model.
Video generation is shifting from one-shot clips to character-consistent production—same face, same voice, controllable scenes. Multi-ref plus native 1080p is the table-stakes bet xAI is making against Runway, Luma, and Google’s Omni-class stacks.
xAI shipped Imagine Video 1.5 with References—a capability update on last month’s Imagine Video 1.5 base. Creators can condition on text, image, and voice (up to seven image refs), run text-to-video without a starter frame, and export native 1080p. Consumer rollout starts with US SuperGrok Heavy and Plus on grok.com/imagine and iOS, expanding to all tiers. API model id grok-imagine-video-1.5 has image refs, text-to-video, and 1080p live; voice reference is available on request.
What shipped#
| Capability | Detail (x.ai primary) |
|---|---|
| Product | Imagine Video 1.5 with References |
| References | Text + image + voice; up to 7 image refs |
| Consistency | Hold face and voice across scenes; swap character vs scene independently |
| Text-to-video | Prompt-only generation (pairs image gen + I2V pipeline) |
| Resolution | Native 1080p for T2V and I2V |
| Consumer access | US SuperGrok Heavy / Plus day one; all tiers rolling out |
| Surfaces | grok.com/imagine, iOS, Android (T2V/1080p called out as GA on those) |
| API model | grok-imagine-video-1.5 |
Prior Imagine Video 1.5 (June) improved motion, physics, and audio. This release is the reference-consistency and resolution step. Distinct from Grok 4.5, Build Mode, and Voice Think Fast 2.0—same Grok Imagine media line, different SKU.
No independent preference-board Elo for this SKU is cited in the primary; quality vs Runway Gen-4.x / Luma remains a buyer-side check.
Limits#
- Voice-ref general availability and API pricing/duration limits are still settling.
- SuperGrok Heavy/Plus first; broader tier timing is rollout, not day-one GA for all.
- Deepfake / identity-leak enforcement is stated as policy, not independently audited here.