Leonardo AI offers a quick prompt box, a more configurable image-creation workspace and an editor for changing an existing image. Start with the route that matches the decision you need to make. Exploring an idea, preserving a reference and repairing one accepted image are different tasks.
If the image is part of an original companion you want to interact with, Cherrypop’s character creator adds personality and scenario choices to appearance. The Leonardo workflow below is useful independently of that option.
Pick the starting route deliberately
| Your situation | Documented route | Decision it helps resolve |
|---|---|---|
| You have an idea, but no model preference | Home-page prompt box | Explore an initial visual direction |
| You need a chosen model, shape or image count | AI Image Creation | Control the generation setup |
| You already have an image worth keeping | Inline editor with a supported Omni model | Request a bounded change instead of restarting |
Leonardo’s generation guide says the home-page prompt can select a model automatically. The image-creation workspace exposes model-dependent settings, including dimensions and image count. The Generate button displays the token cost for the chosen setup. Record that setup before comparing versions: an automatic model switch makes a supposed prompt-only comparison unreliable.
Give each reference a separate job
Consider a cover for a fictional travel journal. You have a sketch showing the composition and an original painted landscape showing the palette. Those references solve different problems. Naming their roles is more useful than uploading several attractive images and asking for a vague mixture.
| Input | Role in this project | Keep out of that role |
|---|---|---|
| Layout sketch | Place the mountain low and leave space above it | The sketch’s rough line quality need not survive |
| Painted landscape | Suggest the muted blue and rust palette | Its buildings and river are not required subjects |
| Written brief | Define the journal cover’s focal point and empty title area | Do not ask the model to invent a final publisher logo |
Leonardo’s current Image Guidance introduction describes contextual reference handling for Omni models. The same page also retains older model guidance. Treat the control shown for your selected model as decisive; a legacy ControlNet instruction is not automatically an available setting in an Omni editor.
Use only the references that carry a necessary constraint. An extra reference is another instruction to reconcile, not evidence of higher accuracy. For this cover, the layout and palette references have clear purposes. A third unrelated poster could make the composition less predictable.
Revise the accepted version, not the entire brief
The Omni and inline-editor guide describes opening Edit on an image, supplying a change instruction and optionally adding references. Model choice affects available inputs. This is a documented editing route; it does not establish that every detail outside your instruction will remain untouched.
- Save the cover version whose composition works. Note its model and settings.
- Describe one change: make the mountain’s lower edge less busy while keeping the skyline and empty title area.
- Generate the revision through the supported editor. Compare the requested area and the surrounding composition.
- Keep the revision only if it solves the problem without introducing a more serious one. Return to the saved baseline when it does not.
This gives a clear stop condition. A cover with an acceptable mountain and readable title space is useful. Repeatedly asking for more detail can consume tokens while filling the space that made the layout work. Add final publication text separately when exact spelling and placement are requirements.
Decide visibility before generating
Leonardo documents a paid Private Mode for keeping generations out of the public feed, with visibility controls discussed in its visibility guide. Check the actual setting before uploading sensitive references. Public-feed visibility is a specific product control; it is not a blanket promise about confidentiality, data retention or every form of provider access.
A fictional, self-owned cover concept is an appropriate first exercise. A confidential client manuscript or unreleased design introduces a separate permission and data-handling decision. Resolve that decision before using it as a convenient test asset.
Check the saved file against its actual purpose
Leonardo documents download, editing and upscaling actions on generated images. Before paying for a higher-cost refinement, inspect the accepted version at the size you intend to use. A thumbnail can conceal edge defects, small accidental symbols and unusable title space.
Keep the source reference, selected model, accepted image and final download together. That record makes a later revision understandable even if the interface or model list changes. Check the file’s actual dimensions and your printer’s requirements before treating it as production-ready.
When the character matters beyond the picture
Leonardo’s workflow suits visual exploration and image editing. Cherrypop is worth considering when the next step is interacting with an original character. Its creator combines appearance, personality and a scenario, then offers routes into chat and media creation. A generated portrait alone does not supply that conversational setup.
Start with the character’s role and opening situation, then choose the visual treatment. Cherrypop is free to start with limits; relevant media may require Premium or Cherries. No identical appearance across all outputs or general layout-editor parity is promised. Create the character and its scenario when that is the project you want to continue.