Someone at a wedding is wearing something wonderful. You spot the bottle on a friend’s dresser, or in a duty-free shop. Could you simply take a photo and know what it is – and how to recreate something close? That is the idea behind the Haveli AI Perfume Decoder.
Step 1: reading the label
The decoder uses optical character recognition to read the text on the bottle and box. It makes several passes – the whole image, the bottle itself, and the small print on the lower label – because perfume labels are often tiny, curved, reflective or written in decorative fonts.
Step 2: matching against a reference library
The text it reads is matched against a reference library of more than 63,000 perfumes. If it can read the brand but not the name, it shows that brand’s perfumes so you can pick the right one in one tap.
Step 3: building the profile
Once identified, the decoder shows the main accords, top, heart and base notes, launch year and mood, and analyses the bottle’s shape and colour from your photo. Then it suggests the closest oils from our library.
Step 4: learning from you
After each result you can confirm it or correct it. Confirmed answers are remembered, so the same bottle is recognised instantly the next time. Photos themselves are not stored – only a fingerprint of the image and the result.
Honest limits
- Heavily stylised or hand-lettered labels can defeat text recognition – typing the name helps.
- Glare, low light and bottles held at an angle reduce accuracy.
- A photo can tell you what a perfume is, not exactly how a specific batch smells.
Tips for the best read
- Face the label straight on, in good light.
- Fill the frame with the bottle.
- If the label is decorative, type the name as well.
Try the Perfume Decoder with a bottle from your own collection.