Grok Imagine Image: 14 aspect ratios and pricing

grok-imagine-image is xAI’s flat-fee image model and Playcut’s budget image tier. It does text-to-image and natural-language image editing. At 10 credits per image it is the cheapest image model in the catalog — Nano Banana 2 costs 34 credits and Nano Banana Pro costs 67 at the same 1K/2K tiers. Pick it for bulk previews, style passes, thumbnail batches, and anything where you will throw away most of the outputs. Do not pick it when you need 4K: this is the only image model we offer that stops at 2K.

Specs at a glance

Model id	grok-imagine-image
Provider	xAI
Modes	text-to-image, reference-to-image (edit)
Aspect ratios	14 — including auto
Resolutions	1K, 2K (no 4K)
Duration	N/A (still image)
Max reference images	2 — xAI’s parser bound is 3, but 3 hits an xAI-side bug; see Limitations
Credits — text-to-image	10 at 1K or 2K
Credits — reference-to-image	20 at 1K or 2K
Multi-region edit	Not supported

Aspect ratios

Grok Imagine Image accepts the widest ratio list of any image model on Playcut, and it is the only one with an auto mode where the model picks the ratio that fits the prompt.

Ratio	Typical use
auto	Model selects the best ratio for your prompt
1:1	Social posts, thumbnails
16:9 · 9:16	Widescreen, mobile, stories
4:3 · 3:4	Presentations, portraits
3:2 · 2:3	Photography
2:1 · 1:2	Banners, headers
19.5:9 · 9:19.5	Modern smartphone displays
20:9 · 9:20	Ultra-wide displays

For comparison, both Nano Banana models support a fixed set of ten ratios (1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9) and neither has an auto option. Grok is the only image model here that offers 2:1, 1:2, 19.5:9, 9:19.5, 20:9, or 9:20 — and the only one that does not offer 4:5, 5:4, or 21:9.

Resolutions

Tier	Grok Imagine Image	Nano Banana 2	Nano Banana Pro
1K	✅ 10 cr	✅ 34 cr	✅ 67 cr
2K	✅ 10 cr	✅ 34 cr	✅ 67 cr
4K	❌ rejected	✅ 43 cr	✅ 84 cr

Requesting 4K on grok-imagine-image is rejected before any credits are held, so a bad request costs you nothing. xAI publishes no pixel dimensions for the 1K or 2K tiers.

Modes

Text-to-image

Prompt in, image out. Nothing else is required.

• Aspect ratio: any of the 14 values above.
• Resolution: 1K or 2K.
• Price: 10 credits per image.

Reference-to-image (edit)

You supply a source image plus a prompt describing the change. There is no @ref or index syntax — you just describe the desired result in plain language.

• Input: one source image.
• Aspect ratio: for a single-image edit the output ratio follows the input image, not the aspectRatio you send.
• Price: 20 credits — 2× the text-to-image price, because xAI bills both the input image and the generated output image.

Reference cap is 2 — verified against the live API

Playcut sends up to 2 reference images per edit (both are used, in the order attached). The cap comes from live probing xAI’s API (2026-07-21): their parser rejects a 4th image outright, and a 3-image request currently fails with an xAI-side parsing bug — so 2 is the reliable maximum, enforced with a clear error before credits are held. If xAI fixes their 3-image handling, this cap will rise.

Pricing

Path	Playcut credits	Notes
Text-to-image, 1K	10	Cheapest image generation on Playcut
Text-to-image, 2K	10	Same price as 1K — no reason to stay at 1K for quality-sensitive work
Reference-to-image, 1K or 2K	20	Input image + output image both billed by xAI
Any 4K request	—	Rejected before the credit hold

Credits are held when the task is submitted and refunded automatically if the task ends FAILED or CANCELLED.

Provider pricing transparency

xAI publishes $0.02 per generated image. It does not publish a separate edit price — it only states that edits bill both the input and the output image. The commonly cited ~$0.04 per edit is arithmetic, not a published number. Our 20-credit edit price is derived from that inference.

Limitations & gotchas

• No 4K. Hard reject at the API layer and again in the worker. There is no silent downgrade.
• Reported maximum source images is contested. One xAI page says up to 5 images for editing; xAI’s own Imagine overview says up to 3. Playcut declares 5. None of these matters today because we send 1.
• Moderation failures produce no image. When xAI’s content filter flags an output, Playcut fails the task rather than delivering the flagged result. The credit hold is refunded, so a moderation hit costs nothing — but you get no output and must rephrase. Grok’s filter trips more easily than Gemini’s on this surface.
• Omitting aspectRatio means auto everywhere. The web picker, API, and MCP all default to auto (the model picks the best ratio for the prompt) since July 2026.
• Not available for multi-region edit. The multi-region edit flow is Nano Banana Pro only.
• The model id is an alias. grok-imagine-image currently resolves to a dated xAI snapshot, so xAI can move the underlying model without a change on our side.

Not selectable through the public API yet

As of 2026-07-21 the public tools registry validates the model field on generate-image-from-text and generate-image-from-reference against a Gemini-only enum, and its aspectRatio enum does not include auto, 2:1, 1:2, 19.5:9, 9:19.5, 20:9, or 9:20. Sending "model": "grok-imagine-image" over the REST API or MCP is rejected by input validation. Grok Imagine Image is reachable from the Playcut web studio today. The example below shows the shape the call will take once the enum is widened.

How to generate

Every Playcut tool uses the same endpoint pattern — see the Quickstart for auth setup and the full task lifecycle.

curl -X POST https://api.playcut.ai/api/v1/tools/generate-image-from-text/execute \
  -H "Authorization: Bearer $PLAYCUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workerId": "65f3a1b2c3d4e5f600000001",
    "prompt": "A neon-lit ramen stall in the rain, cinematic",
    "model": "grok-imagine-image",
    "aspectRatio": "auto",
    "imageSize": "2K"
  }'

The response is the standard async acknowledgement (taskId, status, estimatedCredits). Poll with get-task-status, then fetch the signed URL with get-asset — or receive a webhook. The equivalent MCP tools are generate-image-from-text and generate-image-from-reference.

Related

All models The full Playcut catalog — three image models, seven video models, music and voice — side by side.

Aspect ratios by model Every ratio every Playcut model accepts, side by side.

Resolutions & pricing Credit cost per resolution tier across the whole catalog.

Nano Banana 2 Google's mid-tier image model — 4K capable, 34 credits.

Nano Banana Pro Flagship image model with multi-region editing.