Image generation, sketching, and the image canvas
Connect an image service in Onevium, create images from text, references, or sketches, then compare versions and add annotations for the next edit.
Core capabilitiesTurn an idea into a downloadable image#
Describe a scene in chat, or attach a reference image or sketch. Onevium uses your connected image service to create the result. Inspect it at a larger size, compare versions on the canvas, and add annotations for another edit.
This example makes a coffee-cup illustration.
Connect an image service#
- Open Settings → Providers → Add AI service. Under Image generation, connect a service your account can use.
- After saving, click Image settings beside that entry and choose the Default image model. With multiple connections, choose the default image service.
- Make sure Settings → Tools → Image creation is enabled.
See Providers and models for the fields. Chat models and image services are configured separately. Supported references, sizes, quality levels, and output formats depend on the image model.
Generate your first image#
Send this in the conversation composer:
Create one square illustration for a coffee shop: a blue coffee cup centered
on a cream background, with simple lines and no text. Generate only one image.
When the image appears, check the subject, background, composition, and actual dimensions. If a task returns several images, switch between their thumbnails. Click an image to open the image workspace.
Include a reference image
Paste an image, drag it into the composer, or add it with the file picker. Explain what to keep and what to change:
Use the attached cup as a reference. Keep its shape and blue color,
change the background to cream, and add no text. Generate only one image.
Check the attachment thumbnail before sending. References explain your request; image-input support and reference limits depend on the selected service.
Sketch the composition first#
- Open + → Sketch from the conversation composer.
- Use Pen for outlines and Text for notes. For regular shapes, choose a rectangle or ellipse under Shapes.
- Adjust positions with Select and move. Use Undo / Redo or Eraser to correct a mistake.
- Click ✓ (Attach sketch) at the bottom right, then add “Create a coffee-cup illustration using this sketch's layout” in the composer and send.
The sketch becomes a message attachment. You still need to send the message to start an image task. A rough layout such as “cup in the center, leave space for a title above” is enough.
Compare versions on the image canvas#
Click a generated image to open the workspace, then use the view controls at the top:
| View | Use it to |
|---|---|
| Focus image | Inspect one image closely, switch versions using side thumbnails, and add specific feedback |
| Image canvas | View images from this conversation grouped by generation task; turn on Multi-select to select several |
Select the image you need on the canvas, or double-click it to focus. Zoom changes the viewing scale. Use Resize image, described below, to request different output dimensions.
Add annotations for the next edit#
For example, to remove a small detail on the right of the cup:
- Open the target image, click Comment, and click the point you want to discuss.
- Write the change and click Save comment. The image shows a numbered marker, and the composer keeps the corresponding annotation card.
- Add more feedback if needed. Before sending, check each card's image and instruction; edit or remove annotations you no longer need.
- Add an overall instruction such as “Keep everything else unchanged,” then send and wait for the new result.
The toolbar also provides these editing requests:
| Action | How to use it |
|---|---|
| Mark region | Drag a rectangle around the area to change |
| Erase | Brush over the part to remove; adjust brush size, undo, or clear the marks |
| Remove background | Add a background-removal request for the current image |
| Resize image | Enter the target width and height to add a resize request |
These controls prepare annotations or instructions. The image task runs when you send the message; choosing a tool does not overwrite the original image. Check the new result for both the requested change and the parts you asked to preserve.
Download and inspect the generation details#
- Use Download in the image workspace to save the current result.
- Open Saved prompt and inputs from the more menu to review the prompt and generation options. You can also copy or export the prompt.
- Use Source message to return to the corresponding conversation message.
- Image details in chat shows the model, dimensions, format, elapsed time, and available cost information.
When applicable pricing is available, the cost is an estimate. Cost unavailable does not mean free, and estimates do not replace your provider's bill.
Troubleshooting#
| Symptom | Next step |
|---|---|
| Chat works, but image generation does not | Check the separate image connection, default image model, and Image creation toggle |
| References or dimensions are unsupported | Reduce the references and use options supported by the selected model, then send again |
| The task keeps showing as in progress | Check its status; an animation or estimated progress is not an exact progress report from the provider |
| The outcome is uncertain | Check the request with the provider before generating again, to avoid duplicate requests |
| Images returned, but saving failed | If Retry saving is available, use it to save the images already returned |
| You added marks, but the image did not change | Check the annotation cards, add instructions, and send; marking alone does not run generation |
Next steps#
See Providers and models for connection options, Composer tools for attachments and other inputs, and File preview and editing for local images and files.