xAI has added image and voice references to Grok Imagine Video 1.5, alongside native 1080p output and broader access to text-to-video generation. The update is split between a general release and a narrower early rollout, so availability depends on both the feature and where it is being used.
Text-to-video and native 1080p are generally available through grok.com and the Grok apps on iOS and Android. Image and voice references are starting in the United States for SuperGrok Heavy and SuperGrok Plus customers on the web and iOS, with xAI saying other tiers will follow over the next few days.
Seven images can anchor one scene
The new reference workflow can accept up to seven images in a generation. xAI positions those inputs as anchors for characters, products and locations, giving the model several concrete visual cues rather than relying on a long text description.
That is useful when a video has to preserve more than one element at once. A product demonstration may need a recognisable object, presenter and setting; a narrative scene may need several characters to remain distinguishable. Multiple references do not guarantee that every detail will survive motion, but they provide a clearer specification than a prompt alone.
xAI also shows image and voice references working together. The aim is to retain a character's visual identity while carrying a supplied voice into the generated video. Voice continuity is an important addition because inconsistent sound can break the illusion even when the picture remains convincing.
1080p reaches the app and API
Native 1080p output is now part of the generally available experience. Producing at the target resolution avoids a separate enlargement step, although resolution by itself says nothing about motion accuracy, temporal consistency or the faithfulness of small details.
Developers can use the grok-imagine-video-1.5 model through the xAI API for image-referenced generation, text-to-video and native 1080p output. That brings the main visual capabilities into programmatic workflows such as creative tools, campaign systems or asset pipelines.
Voice references have a different path. xAI says API access is available on request rather than generally enabled. The announcement does not describe the approval criteria, rate limits or safeguards attached to that access, so teams considering a voice-enabled workflow will need to confirm the operational conditions directly.
A rollout with several boundaries
The initial subscription release is limited by geography, plan and client. It begins in the US, applies first to the Heavy and Plus tiers, and launches on grok.com and iOS. Android users have general access to text-to-video and 1080p, but the announcement does not say that reference controls are included in the first Android wave.
Those distinctions are easy to lose in a headline, but they affect whether a workflow can be used immediately. xAI says the reference features will expand to all tiers over the following days, yet it does not provide a precise schedule for every platform or region.
Control is the real product change
The significant part of the release is not merely a larger output file. References give users a way to constrain what the model creates, while 1080p makes a successful result more usable without another processing stage. Together, those changes move Grok Imagine towards repeatable production rather than one-off experimentation.
Important details remain unanswered. xAI's announcement does not publish pricing, generation limits, reference-retention measurements or comparisons showing how reliably identities and voices persist. It also does not spell out consent controls for voice references on the announcement page. The expanded toolset is now visible across consumer and API surfaces; its usefulness will depend on how consistently those controls hold once users move beyond carefully selected examples.