Wan 2.7 Image Pro is the professional tier of Alibaba’s Wanxiang (Wan) image family, released in April 2026 as the upgrade to Wan 2.7 Image. It handles text-to-image generation and prompt-guided editing through the same endpoint, and it’s live on Token Station today as qwen/wan-2.7-image-pro.
What the model does
- Text-to-image at up to 4096x4096. Full 4K output for prompts that need it; 1024x1024 is the default for anything smaller, like a blog thumbnail or a social post.
- Prompt-guided editing at up to 2048x2048. Send an existing image and a new instruction; the model edits it in place.
- Bounding-box editing for precise regions. Add a
bbox_listarray (up to two boxes per image, as absolute pixel coordinates) to point at exactly where an edit should happen instead of relying on the model to find the right region from the prompt alone. - Up to nine reference images. Editing and multi-image generation calls can pull from several inputs at once, useful for keeping a character or a product consistent across a set.
- Text rendering in about a dozen languages. Signs, labels, tables, and simple formulas render with real accuracy, not the garbled text older image models are known for.
- Batch generation. Up to four images per request at the same per-image cost.
- An optional reasoning pass. Wan 2.7 Image Pro can run a “thinking” step before rendering: it works out spatial relationships, composition, and how multiple elements interact before committing to pixels. It helps most on prompts with several interacting subjects or a specific layout, and it adds latency, so treat it as a setting to reach for on complex prompts rather than the default for everything.
What it doesn’t do
- No pixel-mask inpainting. You describe the change in a prompt, optionally aimed at a bounding box; you don’t paint a mask over the exact pixels to replace.
- Character consistency isn’t guaranteed across separate generations. Multi-image reference calls help within one request; two independent calls with the same prompt can still drift.
- Complex multi-section infographics are a weaker fit. For dense layouts with many labeled panels, a model built specifically for that use case will do better.
Generate an image
curl -X POST "https://models.bytefuture.ai/v1/images/generations" \
-H "Authorization: Bearer $YOUR_API_KEY" \
-H "Content-type: application/json" \
-d '{ "model": "qwen/wan-2.7-image-pro", "prompt": "A childrens book drawing of a veterinarian using a stethoscope to listen to the heartbeat of a baby otter." }'$YOUR_API_KEY is your Token Station key, from the dashboard.
Edit an existing image
Add an image_url array with one or more source images and describe the change in prompt:
curl -X POST "https://models.bytefuture.ai/v1/images/generations" \
-H "Authorization: Bearer $YOUR_API_KEY" \
-H "Content-type: application/json" \
-d '{"model": "qwen/wan-2.7-image-pro", "prompt": "Put a trophy in his hand", "image_url": ["https://d3i6fh83elv35t.cloudfront.net/static/2026/06/2026-06-17T033637Z_31440324_UP1EM6H0415VH_RTRMADP_3_SOCCER-WORLDCUP-ARG-DZA-1024x674.jpg"]}'The same endpoint handles both calls. What changes is whether image_url is present, not the route or the model ID.
Parameters
| Field | Notes |
|---|---|
model |
qwen/wan-2.7-image-pro |
prompt |
Required. The generation or edit instruction. |
image_url |
Optional array of source image URLs. Omit for text-to-image; include for editing. Up to nine images per call. |
bbox_list |
Optional, editing only. One list of up to two [x1, y1, x2, y2] pixel boxes per input image, to target where the edit applies. |
Resolution and output
| Mode | Max resolution | Batch |
|---|---|---|
| Text-to-image | 4096x4096 | Up to 4 images per request |
| Editing | 2048x2048 | Up to 4 images per request |
4K is real output, not upscaling, but it’s wasted on anything smaller than a large print or a hero banner. Default to 1024x1024 for web and social use and reserve the higher resolutions for assets that will actually be viewed at that size.
Pricing
Token Station passes provider pricing straight through with no markup. Check the dashboard for the current per-image rate on qwen/wan-2.7-image-pro; other hosts serving the same model price the Pro tier around $0.075 per standard-resolution image, which is a reasonable ballpark while you evaluate it.
Get started
Sign up at models.bytefuture.ai: $1 in free credit, no card required, with up to $50 in bonus credit on your first top-up. Export your key, run the generation call above, then try an edit against one of your own images.