Grok Imagine 2.0
xAI's latest image generation and editing model.
Grok Imagine Image 2.0: AI Image Generation and Editing from xAI
Grok Imagine Image 2.0 is xAI’s image model, and Mitte gives you one tool for both sides of it: type a prompt to generate from scratch, or attach up to 3 images and describe what should change.
Prompts can run up to 8,000 characters, so you can brief it like an art director instead of compressing everything into one line. The model also rewrites your prompt behind the scenes for better results, and you get the revised version back with every generation. Output comes in 13 aspect ratios, from a 2:1 panorama down to a 1:2 vertical, including phone-screen formats like 19.5:9 and 9:20.
What can you do with it?
- Generate images from text. Describe the scene, style, and mood; get up to 4 images per run.
- Edit an existing image with words. Attach a photo and say what changes: “make it night”, “remove the car”, “turn this into watercolor”.
- Combine up to 3 images. Reference several inputs in one edit, like putting a product from one shot into the scene of another.
- Design for phone screens. The 19.5:9, 9:19.5, 20:9, and 9:20 ratios map straight to modern phone displays, handy for wallpapers and app store shots.
- Iterate cheaply, then finish big. Draft at low quality in 1K for $0.04 an image, rerun the keeper at medium quality in 2K.
How to use it
- Write the prompt. What’s in the frame, the style, the light. Long prompts are fine; the model reads up to 8,000 characters.
- Attach images to edit (optional). Drop in up to 3. With attachments, your prompt becomes the edit instruction and the output follows your first image’s ratio by default.
- Pick count and ratio. 1 to 4 images per run, 13 aspect ratios, or Auto to match your input image.
- Set resolution and quality. 1K or 2K, low or medium quality.
- Run it. You get your images ready to download, share, or feed into your next step on Mitte.
Settings
- Number of Images: 1 to 4 per generation.
- Aspect Ratio: 2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20, 1:2, or Auto when editing.
- Resolution: 1K standard or 2K high resolution.
- Quality: low for speed and cost, medium for detail.
- Output Format: JPEG by default, PNG or WebP if you need them.
Pricing
Billed per image: $0.04 (low) or $0.06 (medium) at 1K, and $0.06 (low) or $0.08 (medium) at 2K. Edits add $0.01 per input image. Four medium-quality 1K images cost $0.24.
Tips
- Use the long prompt window. Spell out composition, lens, lighting, and what to avoid. 8,000 characters is room for a full creative brief.
- Read the revised prompt. The model returns the enhanced prompt it actually used. When a result surprises you, that’s where the surprise came from, and it’s a good starting point for the next run.
- Edit in steps. One change per edit pass beats a paragraph of simultaneous changes. Chain edits on the output to stack them.
- Let Auto handle the ratio for edits. It keeps the output matched to your input image, so nothing gets cropped or stretched.
- Generate 4 at a time while exploring. Same prompt, 4 takes, pick the best. At $0.04 a draft image that’s cheaper than fiddling with the prompt one image at a time.