xai-grok-imagine-video

What it does

Guides an AI agent to use xAI's direct Grok Imagine Video API to produce, edit, extend, and govern short videos, handling text-to-video, image-to-video, reference-to-video, and various edits, with attention to pricing, endpoints, authentication, and mode restrictions.

How it works

• Recommends choosing a single active mode per request (text-to-video, image-to-video, reference-to-video, edit, extend) and maps each mode to the corresponding endpoint and parameters.
• Describes endpoint usage, authentication method (Authorization: Bearer $XAI_API_KEY), and the base API URL (api.x.ai). Distinguishes between direct API surfaces and consumer gateways (grok.com, grok on X).
• Enumerates model decisions and allowed inputs/outputs per mode: duration, aspect_ratio, resolution (480p/720p for base; 1080p possible for 1.5), and constraints like input shapes (image, reference_images, video) with explicit valid combinations.
• Details cost structure per model and per output, with example calculations and how to quote costs before approval.
• Provides guidance on production prompts structure (five-part approach) and mode-specific direction for 1.5/base I2V, Reference-to-video, Edit, and Extension.
• Requires preparing a paid boundary dossier before transport, including endpoint/body, price quote, and rights/moderation evidence; recommends dry-run defaults and no automatic retries for creates.
• Includes runnable example outline for a guarded REST client script (xai_video_guard.py) with requirements and behavior, covering all five modes and cost-tracking.

When to use it

Use when you need to orchestrate video generation via xAI Grok Imagine Video API, including:

• Text-to-video, image-to-video, or reference-to-video generation
• Natural-language video edits or extensions
• Asynchronous request handling, batch/files integration, moderation review, and API privacy/residency considerations
• Price-aware approvals with per-second and per-output pricing, and when you must persist intent prior to paid POSTs.

What it can touch

• Endpoints and REST fields for modes (e.g., POST /v1/videos/generations, POST /v1/videos/edits, POST /v1/videos/extensions)
• Model identifiers and accepted inputs: grok-imagine-video, grok-imagine-video-1.5; inputs include prompt, image, reference_images, video, duration, aspect_ratio, and resolution
• Authentication with Authorization: Bearer $XAI_API_KEY
• Pricing data and rate constants used to compute quotes and validate costs
• Example JSON payload shapes shown for reference_images, file_id, and URL-based locators

Caveats

• Clearly separates surface surfaces (grok-imagine-video vs grok on X, grok.com) and notes differences in terms, data controls, and pricing; no API residency promises from pages labeled differently.
• Audio capabilities are provider-claimed; the video REST schema does not document an audio enable/disable field or related audio parameters.
• Accepts potential documentation conflicts on media size limits (48 MB vs 50 MB) and advises staying below 48 MB or verify exact endpoint before upload.
• First-party rate limits are noted as UNKNOWN from public docs; team-specific Console limits should be used.
• Requires careful handling of sensitive data, logs, and private file locators; avoid logging Base64 or signed URLs and apply strict access controls.