xAI Studio

xAI Studio — generate and edit images and videos via the xAI API. Image: text-to-image, batch generation, multi-image editing, concurrent style transfers, mu...

Security Scan

The skill's code, declared requirements, and runtime instructions are consistent with an xAI image/video CLI: it requires only python and an XAI_API_KEY, installs/uses the xai-sdk, and reads/writes media files locally.

Purpose & Capability

Name/description (xAI Studio) matches the actual behavior: generating and editing images/videos through the xai_sdk. Required binary (python3) and primary env var (XAI_API_KEY) are proportionate to the described purpose.

Instruction Scope

SKILL.md and scripts/run.py confine actions to using the xai_sdk, encoding local files, optionally downloading user-supplied image/video URLs, and saving outputs under media/xai-output. This is expected, but note that user-supplied URLs will be fetched (urllib.urlretrieve) and local files may be read and base64-encoded and sent to the xAI service.

Install Mechanism

No high-risk installers are present. SKILL.md recommends creating a venv and pip-installing xai-sdk; metadata includes apt/pip steps. Small inconsistency: the registry summary said 'No install spec — instruction-only', yet SKILL.md contains install entries and a code file is included. The pip package xai-sdk is the expected dependency; verify its provenance before installing.

Credentials

Only XAI_API_KEY is required as the primary credential, which is appropriate for calling the xAI API. The code does not read other environment variables or system config paths.

Persistence & Privilege

always is false and the skill does not request persistent system privileges or modify other skills. It writes outputs to a local media directory only (no system-wide changes).

Assessment

This skill appears coherent for its stated purpose, but consider these before installing: - The xai-sdk will send your provided images/videos and prompts to xAI; do not upload sensitive/private media unless you accept that it will be sent to the service. - The skill requires your XAI_API_KEY; treat it like any API secret. Only provide a key with the appropriate scope and rotate it if you stop using the skill. - SKILL.md suggests installing the xai-sdk via pip. Verify the xai-sdk package identity (official repository or PyPI publisher) before installing in a shared environment. - The CLI will fetch any image/video URLs you pass (urllib.urlretrieve). Only pass trusted URLs to avoid unexpected network activity or malicious payloads. - Minor metadata inconsistency: the registry listing noted 'instruction-only' but the package includes a scripts/run.py and install instructions. That’s not dangerous, but be aware the skill includes executable Python code that will run locally.

Generate and edit images and videos using xAI's models via the xAI SDK.

Setup (Venv + SDK)

Create an isolated virtual environment and install the SDK:

python3 -m venv venv
source venv/bin/activate
pip3 install xai-sdk
deactivate

Generate Images

venv/bin/python3 scripts/run.py generate --prompt "A futuristic cityscape at dawn"

Generate multiple images in one request:

venv/bin/python3 scripts/run.py generate --prompt "Abstract art" --count 4

Edit Images

Edit an existing image with a text prompt:

venv/bin/python3 scripts/run.py edit --prompt "Make it a watercolor painting" --image photo.png

Combine up to 3 source images:

venv/bin/python3 scripts/run.py edit --prompt "Add the cat to the landscape" --image cat.png --image landscape.png

Concurrent Style Transfers

Apply multiple styles to the same image in parallel:

venv/bin/python3 scripts/run.py concurrent --image photo.png --prompt "oil painting" --prompt "pencil sketch" --prompt "pop art"

Multi-Turn Editing

Chain edits sequentially — each output feeds into the next:

venv/bin/python3 scripts/run.py multi-turn --image photo.png --prompt "Add dramatic clouds" --prompt "Make it a sunset"

Common Flags

• --model: Model name (default: grok-imagine-image)
• --aspect-ratio: e.g. 16:9, 4:3, auto (default: 1:1)
• --resolution: 1k or 2k (default: API default)
• --format: base64 (default) or url
• --out-dir: Output directory (default: media/xai-output)

Output

Images are saved to <out-dir>/<YYYY-MM-DD>/<prefix>_<NNN>_<HHMMSS>.<ext>, organized by UTC date. The prefix reflects the subcommand: generate, edit, style (concurrent), or step (multi-turn). The file extension is detected automatically from image magic bytes (PNG, JPEG, WebP, GIF).

---

Video: Generate

# Text-to-video
venv/bin/python3 scripts/run.py video-generate --prompt "A rocket launching from Mars" --duration 10 --resolution 720p

# Image-to-video
venv/bin/python3 scripts/run.py video-generate --prompt "Animate this scene" --image photo.png --duration 5

Video: Edit

venv/bin/python3 scripts/run.py video-edit --prompt "Add a silver necklace" --video https://example.com/clip.mp4

Video: Concurrent Edits

venv/bin/python3 scripts/run.py video-concurrent --video https://example.com/clip.mp4 --prompt "Add a hat" --prompt "Change outfit to red"

Video Flags

• --model: Model name (default: grok-imagine-video)
• --duration: Length in seconds, 1–15 (default: 5) — generate only
• --aspect-ratio: e.g. 16:9 (default: API default)
• --resolution: 480p or 720p (default: API default)
• --timeout: Max polling wait in seconds (default: SDK default)
• --poll-interval: Seconds between status checks (default: SDK default)

Videos are saved as .mp4 to <out-dir>/<YYYY-MM-DD>/<prefix>_<NNN>_<HHMMSS>.mp4. The prefix reflects the subcommand: video (generate), video_edit, or video_style (concurrent).