Sylvester uses computer vision and recognized feline pain science to help you understand how your cat is feeling, right from your phone
Comfort insights built on the same science veterinary professionals use
Cats are wired to hide pain. Unlike dogs, they rarely vocalize or limp when something is wrong. Instead, discomfort often shows up as subtle shifts in facial expression such as a slight tension around the muzzle, ears that sit a little lower, eyes that aren't quite as open as usual.
These signs are easy to miss, even for experienced cat caregivers. And because cats tend to mask how they feel, many are living with discomfort that goes unnoticed.
The science behind the scores
Sylvester is built on a dataset of hundreds of thousands of real cat images, sourced from veterinary clinics, shelters, and caregiver submissions from around the world. Every image was expert-annotated using validated feline pain assessment frameworks, the same science veterinary professionals rely on to evaluate discomfort in cats.
There are three fully validated acute pain assessment tools for cats, and Sylvester draws on all three.
Feline Grimace Scale
FGS or Feline Grimace Scale was developed at the University of Montreal,and scores five specific facial markers. Each marker is scored on a scale, and there is a specific threshold above which the science indicates a cat needs attention.
Glasgow Composite Measure Pain Scale
Developed at the University of Glasgow, evaluates both facial expression and behavior and has been validated across a wide range of pain types including post-surgical recovery, trauma, and medical conditions, and includes a threshold at which care is needed.
UNESP-Botucatu Multidimensional Composite Pain Scale
Developed at São Paulo State University in Brazil, is the most extensively validated of the three. It has been tested across multiple languages, pain types, and clinical settings, making it one of the most thoroughly studied tools in feline pain science.
Until now, applying any of these frameworks has required a trained professional in the room with your cat. Sylvester brings that same science into every home.
Sylvester was built in close partnership with feline specialists, veterinary clinics, and shelter partners, with our model trained to recognize the subtle facial signals that indicate discomfort across hundreds of thousands of cats.
Why general AI can't accurately detect feline discomfort
General AI tools including large language models and general image recognition systems aren't trained on medically labeled datasets of cats. They haven't been built around validated feline pain science, and they haven't been taught to distinguish the subtle facial shifts that signal discomfort in a species that instinctively hides it.
Detecting feline pain requires a purpose-built model trained on the right data, labeled by experts, using the right frameworks.
That's not something a general AI can approximate. It's something that has to be built from the ground up — which is exactly what Sylvester has done.
Read our Technical Whitepaper
See the science behind Sylvester.ai, from our image dataset to the veterinary frameworks and validation that power our technology.
How Sylvester Decodes What LLM’s and General AI Can’t
Accurately detecting feline pain requires specialized datasets of real cats with professionally labeled facial expressions and health outcomes — data that doesn't exist anywhere else.
Read What Caregivers Say
Hear from caregivers and veterinary teams sharing real stories of discovery, peace of mind, and more collaborative care.
A cat comfort check in three simple steps:
What Caregivers are saying about Sylvester
“We are investing in sylvester.ai as a differentiator, but honestly, every clinic in town should be using it.” - Investor and Veterinary Clinic
“We are investing in sylvester.ai as a differentiator, but honestly, every clinic in town should be using it.” - Investor and Veterinary Clinic
I had no idea if I was taking proper care of my cat or not. When the app told me he was unhappy, I decided to take him to the vet and he was diagnosed with ear mites. ...Many thanks to sylvester.ai which provoked me to take my cat to the vet and therefore saved him a lot of pain." - Cat Caregiver
I had no idea if I was taking proper care of my cat or not. When the app told me he was unhappy, I decided to take him to the vet and he was diagnosed with ear mites. ...Many thanks to sylvester.ai which provoked me to take my cat to the vet and therefore saved him a lot of pain." - Cat Caregiver
“I'm glad I used the app to check on my cat or I wouldn't have learned his ear infection was still present!” - Cat Caregiver
“I'm glad I used the app to check on my cat or I wouldn't have learned his ear infection was still present!” - Cat Caregiver
“This app saved my cat’s life.” - Cat Caregiver
“This app saved my cat’s life.” - Cat Caregiver
Tips for comfort checks
AI is a powerful tool, but it is not perfect. Sylvester.ai is not a diagnostic test. It is a support tool that helps cat caregivers notice signs of pain. To make sure the results are reliable, photos need to meet some basic requirements. If you ever have concerns about your cat’s health, always consult your veterinarian.
-
Age
Best results come from adult cats. Kittens often produce less reliable scores.
-
Breed Type
Works best with average facial structures. Cats with extreme features such as flat-faced (Persian, Himalayan) or long-nosed (Abyssinian, Siamese, Sphynx) may be less accurate.
-
Facial Visibility
The cat’s face should be fully visible. Covered or partially blocked faces cannot be assessed.
-
Angle
Cat should be facing forward with the full face visible. Side views or turned heads reduce accuracy.
-
Lighting
Photos should be taken in bright, even light. Dark, shadowed, or backlit photos make features harder to detect.
-
Sharpness
Images must be clear and in focus. Blurry or pixelated photos cannot be scored accurately.
FAQs
-
Sylvester.ai is not generative. It does not invent or imagine results. Instead, it is trained on expert-labeled cat images and applies repeatable, science-based rules to identify pain cues. This makes it grounded in veterinary research rather than creative output.
-
“Not happy” indicates that Sylvester.ai detected signs of discomfort or pain. It does not diagnose a condition but signals that a closer look or a vet visit may be needed.
-
If the image is too dark, blurry, or not facing forward, Sylvester.ai will ask you to try again. If you keep running into issues, contact us at support@sylvester.ai.
-
If something looks wrong, please contact us at support@sylvester.ai and our PhD animal behaviorist will review your feedback and photo. If you suspect your cat may be in serious discomfort, keep in touch with your veterinarian for further care.
Key Research Papers
Investigating the capabilities of large vision language models in dog emotion recognition. Sci. Rep. 15: 41250. Martvel et al. 2025. https://doi.org/10.1038/s41598-025-25199-7
Evaluation of facial expression in acute pain in cats. J. Small Anim. Pract. 55: 615-621. Holden et al 2014. https://doi.org/10.1111/jsap.12283
Behavioural signs of pain in cats: an expert consensus. PLoS ONE 11(2): e0150040. Merola & Mills. 2016. https://doi.org/10.1371/journal.pone.0150040
Geometric morphometrics for the study of facial expressions in non-human animals, using the domestic cat as an exemplar. Sci. Rep. 9: 9883. Finka et al 2019 https://doi.org/10.1038/s41598-019-46330-5
Development of a behavior-based measurement tool with defined intervention level for assessing acute pain in cats. J. Sm. Anim. Pract. 55(12): 622–629. Calvo et al 2014. https://doi.org/10.1111/jsap.12280
Validation of the English version of the UNESP-Botucatu multidimensional composite pain scale for assessing postoperative pain in cats. BMC Vet. Res. 9: 143 – 158. Brondani et al 2013. https://doi.org/10.1186/1746-6148-9-143
Facial Expression: An Under-Utilized Tool for the Assessment of Welfare in Mammals. ALTEX 34 (3) :409-429. Descovich et al 2017. https://doi.org/10.14573/altex.1607161
Definitive Glasgow acute pain scale for cats: validation and intervention level. Vet Rec. 108(18): 444-446. Reid et al 2017. https://doi.org/10.1136/vr.104208