Creative testing
Compare campaign images early and focus production on the strongest direction.
Introducing Cwupid 1.1
Created by our team of computer vision engineers, Cwupid 1.1 sees the whole photo, not a made-up face formula, to help you choose and share your most engaging photos.
01
Choosing is the part nobody is good at. You have been looking at the same nine shots for ten minutes, and the longer you look the less you can tell them apart. Ask a friend and you get a kind answer. Post it and you find out too late.
Cwupid is the outside opinion, in about ten seconds.
02
One shot, or a whole set from the weekend. JPEG, PNG or WebP, straight off your phone. Your face does not have to be visible.
Framing, light, setting, styling, and the way it all comes together — the things that decide whether a photo lands, most of which are not your face.
A percentile against real posts, plus a short written read of what is working and what you could change. Then you pick, and post.
03
A formula built on symmetry and proportions gives you the same answer for every photo you are in. That is not how photos work. The good one and the bad one from the same evening have the same face in them.
Cwupid was built to read what actually differs between the two.
04
94 means Cwupid read this photo ahead of 94 out of every 100 comparable posts it has seen. It is a reading of how an image comes across — not a measurement of you, and not a verdict on anybody in it.
Cwupid learned from 1.6 million real-world images: mirror selfies, graduations, group shots, bad lighting, actual life. That is the bar your photo is being read against.
05
Photo Studio takes a whole set, scores every frame, and hands back the ones worth posting — either your strongest few, or a carousel sequenced so it opens strong and holds together.
Your best photo is already in the camera roll. This is the fastest way to find it.
Use Cwupid to compare visual creative, understand what draws attention, and choose stronger images before you publish.
Compare campaign images early and focus production on the strongest direction.
Help social, brand, and growth teams choose images with clearer audience appeal.
See which parts of an image carry the most weight, then turn that into a repeatable creative playbook.
For pilots, integrations, or business access, get in touch.
cwupid@cwupid.comCwupid is a small team building a photo model and the product around it. Two roles are open. Both are hands-on from the first week, because there is nobody to hand the work to.
Engineering
You own the thing customers actually touch: the site, the signed-in app, billing, and the edge service behind them. It is a small codebase with real users and real money moving through it, so the bar is care rather than volume.
Nice to have: Cloudflare D1 or R2, durable queues, Python for operational tooling, or having been the person on call.
Machine learning
You own the model and the machine it runs on. Cwupid scores a photograph end to end in seconds on modest hardware, and every improvement has to survive that constraint as well as the evaluation.
Nice to have: AWS (EC2, Lambda), ONNX or quantised inference, integrated gradients and occlusion methods, or experience with human-subject data policy.
Email cwupid@cwupid.com with the role in the subject line, a short note on something you have built, and a link to your work. No cover letter. We reply to everyone.
Cwupid 1.1 is made for the question you actually have: which photo will make the best impression?
Cwupid 1.1
Most moments are not taken under perfect lighting, at the perfect angle, with the perfect background. Cwupid understands the entire image, so its read feels closer to how a photo actually comes across.
A great photo is more than symmetry, proportions, or a checklist of facial features. Cwupid looks for the feeling of the photo: presence, style, energy, composition, and the little things that make someone pause.
From mirror selfies to graduation shots to your next carousel, Cwupid was built around the varied, in-the-wild images people share every day.
A better way to choose
Let Cwupid help you find it.
Cwupid was trained from 1.6 million images. Your face does not have to be visible. Add a portrait, profile, full-body shot, or scene-led photo.
Score the same crop separately with selected learned college-page embeddings. All other reference-context fields remain fixed.
Drop a photo here, or pick one from your device.
Unsaved photos are discarded after scoring. Sign in to save a photo and its report to your private history.
The live scorer runs on a single machine which is not running right now. The recording below shows the same path a live upload takes.
Cwupid's range for this photo ….
Sign in to keep this result in your private history.
Add up to 20. Cwupid scores every one, ranks the set, and only writes notes for your winners.
Drop your photo set here
JPEG, PNG, or WebP. Up to 20 photos.
Unsaved photos are discarded after this result. Account storage currently saves individual scored-photo reports.
Cwupid is ranking every photo before it writes a single note.