Getting started with BANANAI

Start with your goal

Goal	Where to start
Ask questions, write, or analyze an image	Create an AI chat and add text or attachments supported by the model.
Generate an image from text or references	Create an AI image session, then enter a prompt and settings in Canvas or Cards.
Split, crop, rotate, or annotate an image	Select an image in the image workspace and open Image editing.
Extend a scene or change a local area	Use Outpaint in the image tools, or mark an area through the generation reference-image editor.

Try an exercise without generation credits

Create an AI image session. Prepare a practice image without private information. Select it through the canvas upload control, confirm, and wait for upload success. It becomes a standalone image node.
Open Transform for that image and try cropping or rotation. Check the preview dimensions before applying. Transform does not consume generation credits, but its new result needs storage.
Inspect the canvas: the original remains, alongside an operation and result. Connections show where the result came from. This is not an overwritten original or a failed AI generation.
Switch to Cards to see the same editing record with input, tool, and output together. Switching back retains the same relationship; it does not generate again or charge again.
Open the result, download its original, and check the file. Favorite it to return to this step. To keep its creative method, check the upstream scope and collect it as inspiration.

**Transform preview and execution are separate**
Check crop and dimensions before applying. Switching tools or preview settings does not overwrite the original.
Actual interface · Tutorial example, not your account

**Real editing results retain their sources**
Horizontal mirror produces one result, followed by annotation. Connections show direct relationships and each original input remains.
Actual interface · Tutorial example, not your account

Generate your first image

Choose Start a new creation in the sidebar or empty workspace, then AI images - Infinite canvas. With no active image view, it opens in Canvas.
Open generation through the canvas prompt-node control, or switch to Cards and add an AI image draft at the end.
Describe the subject, scene, style, and details to retain or exclude. Select references if needed and wait for upload success.
Choose an available model and check its actual aspect-ratio or size, quality, and other controls. Check reference count, credit estimate, and permission messages.
Submit. Queued or generating does not mean failed. When complete, open the image, download the original, or continue creating.

**Actual generation settings before submission**
Check references, prompt, model, and its supported settings. This is a filled draft, not a claimed successful image. Credit estimates may appear separately during submission; current settings and prices take precedence.
Actual interface · Tutorial example, not your account

Turn generation into an understandable process

Test one clear goal first: for example, keep the white vase and plant, but replace the red cube with a blue sphere. Mark a local area when appropriate. Do not ask to preserve everything and change the whole composition without specifying priorities.

Use Continue creating to develop a chosen result into a new step. Use Regenerate for another candidate from the same operation settings. These actions have different relationships, and neither guarantees a free repeat.

Check in this order: correct inputs, consistent prompt, suitable settings and credits, requirements met by the result, then keep, continue, or branch. This makes problems easier to identify than repeatedly generating.

What is saved

Created sessions, formal nodes, and generation results belong to the workspace. Cards and Canvas show the same content; switching does not require generation.

Unsubmitted prompts, drawings, and some panel layouts are workspace drafts. Complete any operation you need to keep before refreshing or closing; a draft is not a saved result.

Plans, available models, image counts, storage, and prices depend on the current account and site configuration. Follow the operation page and submission result.