Bring the visual direction. Generate a new layer structure.
Upload an image as a compositional reference, explain the result you want, and let LayerAxon generate the requested elements as independent layers. The reference guides the new work; it is not treated as a flat file to cut apart.
Reference guidance
Composition, subject placement, palette, and visual character
The uploaded image provides context. It does not reveal the source document or any layers that existed before it was flattened.
Generated layer plan
Illustrative names only. The real layer plan follows your prompt, industry, target count, and selected settings.
Reference is a brief, not a hidden source file
A JPEG or PNG shows the visible composite. It does not contain the original object hierarchy, masks, off-canvas pixels, type settings, or layer names. A workflow that claims to recover those exact originals would be promising information the image does not hold.
LayerAxon takes a different route. The reference helps establish what the new composition should feel like and where major elements belong. Your prompt and layer direction describe the intended outcome. The system then generates new visual layers and assembles them into an editable layered file.
A practical reference-led workflow
Choose what the image should guide
Decide whether the useful signal is layout, palette, subject placement, atmosphere, or a combination. A clear intention is more useful than asking for a vague copy.
Upload the reference or write from text
The Layered AI input supports either an uploaded visual or a text-first request. With a reference attached, your written direction can focus on what should remain recognizable and what may change.
Define the elements you need later
Set the industry, target layer count, quality, style, and output format. Use layer direction to name the controllable parts that matter in the downstream edit.
Review the new interpretation
Inspect the generated preview for composition, boundaries, and useful separation. Confirm the result only when the layer structure is a sound starting point for the next edit.
Good reference-led requests
- Keep the low camera angle and warm backlight, but generate the product, surface, shadow, and background independently.
- Use this architectural massing as the composition guide; create separate sky, building, landscape, people, and atmosphere layers.
- Preserve the central silhouette and color relationship while generating a new concept-art scene with editable foreground and environment.
Requests that need reframing
- “Recover the exact original PSD.” A flattened reference cannot supply information that is no longer present.
- “Make every detail identical.” Generation can follow guidance, but it is not deterministic forensic reconstruction.
- “Make it editable.” Name the elements you intend to move, hide, replace, or revise so the layer plan has a concrete purpose.
Reference-image questions
Does the reference image get split into its original layers?
No. A flat image does not contain its original hidden layer structure. LayerAxon uses the image as guidance while generating a new set of visual layers.
What should a reference image communicate?
It can communicate composition, subject placement, color relationships, or an overall visual direction. Your text instructions and layer plan clarify what the new output should preserve or change.
Will the generated PSD match the reference exactly?
Not necessarily. The workflow generates a new interpretation rather than reconstructing the source file. Review boundaries, details, and consistency before production use.
Use the image to direct what comes next.
Open Layered AI with a prompt or a visual reference.
Prefer a text-only brief? See text to layered PSD.