Grok Imagine

Positioning Overview

grok-imagine/image-to-video is an image-to-video model in xAI's grok-imagine family, called through the unified HiAPI.ai interface. It takes one or more reference images together with a text prompt and turns static frames into moving shots, supporting durations of 6–30 seconds, resolutions from 480p to 1080p, and multiple aspect ratios. It suits developers and creative teams that need to quickly animate illustrations, storyboards, or product images for social assets, ad prototypes, and motion-preview shots.

Key Highlights

• Multi-image reference guidance for motion — you can pass up to 7 reference images and reference specific ones in the prompt with @image1 and similar tags, making it easy to finely control the subject and composition.
• Switchable motion styles — three generation modes are available: fun, normal, and spicy, corresponding respectively to more creative, balanced, and more dynamic motion.
• Duration and resolution can be balanced to your needs — duration can be set between 6 and 30 seconds, and resolution covers 480p to 1080p, making it easy to trade off generation speed against final video quality.
• Clear billing basis — billing is by output seconds, and the duration parameter directly determines the cost of a call, making it easy to estimate budgets before batch generation.

Capabilities

Reference-image-driven video generation: input 1–7 JPEG, PNG, or WEBP images, and the model generates moving footage while preserving the subject and composition, making it suitable for turning static assets into shots.

Text prompts control motion: you can optionally pass a prompt of up to 5000 characters describing the desired motion; English descriptions work better, helping the model focus on the specified actions and camera direction.

Multi-aspect-ratio output: supports five ratios: 2:3, 3:2, 1:1, 16:9, and 9:16, effective in multi-image reference mode, covering both landscape and portrait placement needs at once.

Adjustable resolution and duration: output resolution defaults to 480p and can be raised to 720p or 1080p; duration defaults to 6 seconds and can reach up to 30 seconds, billed by the actual number of generated seconds.

Use Cases

• Designers animating static illustrations into motion shots — use an illustration or storyboard frame as input with a short motion prompt to produce a short video shot for animated posters, title sequences, and storyboard previews.
• E-commerce teams generating dynamic product displays — upload a product image and describe the camera motion to get a portrait or square short video that can be used directly on product detail pages and in social feed placements.
• Content creators producing social assets in batches — generate motion assets in batches at 9:16 and shorter durations for daily content updates and creative testing on short-video platforms.

Limitations and Selection Advice

It only does image-to-video and does not accept pure text input to generate video, so at least one reference image is required. With a single image input, the aspect ratio parameter has no effect, and the output follows the original image's ratio; 1080p resolution only supports a single input image. When using external images, the generation mode may be automatically downgraded to normal. If you need to generate video directly from text, or need continuous long shots longer than 30 seconds, choose another model or generate in segments. Cost grows with duration, so it is advisable to run a small sample test before batch calls with long durations.