A clear answer.
Your next step.

Choose Create character, select a reference image first, then adjust and confirm it once before choosing a style in the workspace. Naming is optional when starting; a name is required when saving to My characters. Review the operation and credit cost before submitting.

The current workspace accepts JPG, PNG and WebP up to 10 MiB (10,485,760 bytes) and at most 8 megapixels. Start with one clearly visible subject you have permission to edit. A distant, blurry or hidden face gives you fewer details to compare.

The current creation flow starts with one reference image. It does not offer text-only generation, multiple-reference blending or personal model training. Use one clear subject, then describe the change you want. For later creations, reopen a saved character to reuse its main reference.

Choose a style preset for the rendering medium, then describe the specific outfit, expression or setting you want. Start with a small change and compare the output with your reference. For example, try a cream sweater in a softly lit library. This is a suggested prompt, not a recorded result or a guarantee of an exact pose or likeness.

Cancelling the file picker does not select a new file. Cancelling a replacement in the crop dialog keeps the previous workspace reference. The new reference replaces it only after the crop is confirmed. Closing the separate new-character dialog clears that dialog’s draft.

Choose the medium in the workspace style gallery. Open More options for the creation type. Expressions requests six editable default expressions; Character sheet requests front, side and back images; Scene set requests four scenes. You can edit the default expression, view and scene descriptions.

Save a version as the character's main reference, then reopen that character for later requests. Compare each scene or expression with the saved reference before judging the new setting or mood. Edit uses the selected result as input, so return to the main reference if a version drifts. Exact consistency is not guaranteed.

Reopen the saved character so each attempt starts from its main reference. Describe one new setting in A few details, or use Scene set for four descriptions. Review the character and the background in every result; a new scene can also change visible facial details.

A new account receives a one-time allowance of 4 credits, valid for 30 days. That covers one standard single image at 4 credits, not unlimited free images. It does not cover a 24-credit six-expression set or a 12-credit three-view set.

Single image: 4. Edit: 4. Six expressions: 24, at 4 per image. Three character views: 12, at 4 per image. Four scenes: 16. Background removal: 4. Upscale to 2K: 12. Check the selected operation before submitting.

Each successful child image consumes its reserved share. A confirmed failed child releases its own reservation. For example, five successful expressions and one confirmed failure consume 20 credits and release 4. This example is arithmetic, not a recorded generation result.

Not necessarily. A provider or settlement result that cannot yet be confirmed may keep credits reserved while recovery or status checks are needed. Use Check task status before submitting another job. Contact support with the task ID and visible status; do not treat an unresolved task as a confirmed failure.

Yes. Open the successful creations, check the full image and download the ones you want to keep. Downloading an existing result does not consume generation credits. A new generation is a separate job with its own credit cost.

Uploads and creations are private to your workspace and are not automatically published in a public gallery. Public homepage examples are separately labelled concept artwork and approved examples.

Generated versions remain accessible in the workspace for 30 days. This access period does not mean the files are automatically physically deleted from Cloudflare storage. Saved main references depend on your character-space allowance. After a downgrade, extra spaces enter a 30-day read-only period. Download images you want to keep.

Morrow uses Kie.ai for image processing, Waffo Prompt Sift for prompt screening, and SeeAPI for additional safety review. Prompts, reference images and generated images are sent for the relevant processing or review. We do not make a zero-training or permanent-deletion promise about these providers.

No. Check the source material, likeness permission, supplier and model terms, and the intended use. A paid Morrow plan by itself does not grant blanket commercial rights, establish copyright ownership or remove provider restrictions.

No. The outputs are 2D images. Character sheet requests front, side and back images, not a rotatable mesh, rig or editable geometry. Hidden details absent from the reference can be inferred and need human review.