AI & Accuracy Disclaimer
Last updated: 28 May 2026
COVERvision uses artificial intelligence to generate visual previews of furniture with different fabrics and design features. This page explains how our AI works, what it can and cannot do, and how to get the most from it.
Important: Previews generated by COVERvision are AI-generated and intended as a visual guide only. They should not be relied upon as a definitive representation of the finished upholstery result. We always recommend ordering physical fabric samples before making any purchasing or project decisions.
How COVERvision AI works
COVERvision uses a large language model (Claude, by Anthropic) with image generation capabilities to analyse your furniture photograph and fabric swatch, then generate a new image showing how that fabric might look on the furniture.
The AI attempts to:
- Apply the colour and texture of the fabric swatch to the upholstered surfaces in the photo
- Preserve the shape, dimensions, and non-fabric elements of the original furniture
- Incorporate design changes specified in your Design Notes (e.g. piping, tufting, button details)
- Maintain consistent lighting and perspective from the original photograph
What AI visualisation does well
Showing a clear colour change — how a sofa looks in navy vs cream vs rust
Demonstrating fabric texture changes — from plain to velvet, leather to linen
Adding design features that aren't on the original piece — piping, nailheads, button tufting
Giving a strong directional sense of whether a fabric "works" with a piece
Comparing multiple fabric options across saved previews
Known limitations
Colour accuracy
Screen calibration, photo lighting, and AI interpretation can all affect how a colour renders. The generated image may not exactly match the fabric's true colour. Always request a physical cutting to confirm colour before proceeding.
Pattern scale and repeat
The AI may not accurately reproduce the scale of a patterned fabric. A large-repeat pattern may render smaller or larger than it would appear in reality.
Texture depth
High-pile fabrics like velvet, boucle, or heavy weaves may not show their full depth and texture in a flat swatch photo. Real-world texture can differ significantly from a 2D representation.
Existing design features
Changing or removing existing design elements from the original furniture photo (e.g. existing piping, buttons, or pleats) can be inconsistent. The AI works best when adding features rather than removing them.
Complex furniture shapes
Highly curved, tufted, or multi-section furniture can be more challenging to visualise accurately. Results on simpler, well-lit pieces tend to be more reliable.
Photo quality dependency
The quality of the output is directly influenced by the quality of the input. Poor lighting, blurry images, or partial views of the furniture will produce less accurate results.
Tips for the best results
- Use official fabric swatch images from your supplier's website — they're photographed in controlled lighting
- Photograph your furniture in good natural light, straight-on, with the full piece visible
- Be specific in your Design Notes — use directional language like "ADD piping on all seams" or "REMOVE buttons, smooth finish"
- If the result isn't right, use the Suggest Improvement button to refine rather than starting over — it uses your previous result as a starting point
- Always order a physical fabric cutting before confirming any fabric or design decision
Third-party AI provider
COVERvision uses the Claude API provided by Anthropic, PBC for AI image processing. Your uploaded images are transmitted to Anthropic's API for processing in accordance with Anthropic's Privacy Policy. We do not use your images to train AI models without your consent.
No guarantee of physical outcomes
COVERvision previews are design exploration tools. They do not constitute a guarantee, quote, or specification for upholstery work. Any decisions about physical upholstery projects should be made in consultation with your upholsterer, with reference to physical fabric samples and professional assessment.
Questions
If you have questions about our AI technology or accuracy, contact us at support@covervision.co.nz.