xAI Studio#

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:

bash

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

Generate Images#

bash

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

Generate multiple images in one request:

bash

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

Edit Images#

Edit an existing image with a text prompt:

bash

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

Combine up to 3 source images:

bash

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:

bash

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:

bash

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#

bash

# 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#

bash

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

Video: Concurrent Edits#

bash

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).