Animal Vision

Methodology and sources

Last updated: September 7, 2026

Animal Vision renders what 46 animals see, live through a phone camera. Every number it shows — cone sensitivities, visual acuity, field of view, flicker rate — is taken from published vision research rather than estimated. This page explains exactly how a simulation is built, what it can and cannot tell you, and lists every source behind it.

How a simulation is built

1. The research comes first

A species is only added once we can find primary literature that measures its vision: photoreceptor spectral sensitivity, the number of cone types, behavioural or electrophysiological colour-discrimination evidence, and an acuity figure with units. If those numbers do not exist, or the studies disagree too much to pick a defensible value, the species is not added. Every figure is recorded with its source, its method, its sample and its uncertainty before a single line of rendering code is written.

2. Colour

Each species' cone peaks determine how its retina divides the spectrum. Frames are converted from sRGB into linear light, transformed with a matrix derived for that receptor set, and converted back. For dichromats — most mammals, including dogs and cats — the transform follows the widely used Machado, Oliveira & Fernandes (2009) model of altered colour perception. Species with a single receptor class collapse to luminance with a species-appropriate tint instead.

3. Sharpness

Acuity is published in cycles per degree (cpd): how many alternating light-and-dark stripes an eye can still tell apart within one degree of visual angle. A human reaches roughly 60 cpd; a dog about 12; an eagle about 140. We convert that figure into the blur radius that removes exactly the detail the species cannot resolve, so the softness you see is a measurement rather than an artistic choice. That is also why most species look blurry — it is the finding, not a defect.

Each species page also gives an eye-chart (Snellen) score. We use one convention for all of them, the clinical one, in which 20/20 corresponds to 30 cpd: a species resolving c cpd is shown as 20/(600 ÷ c), rounded. On that scale a young adult's best acuity of about 60 cpd is 20/10, a dog's 12 cpd is 20/50 and an eagle's 140 cpd is about 20/4. Where a study quotes its own Snellen figure, the article names it as that study's estimate rather than a conversion.

4. The same code renders the app and this site

The colour maths lives in one framework-free module that both the mobile app and this website import. The colour comparisons printed on every species article are computed at build time by the identical function that transforms the live camera frame, so the site can never illustrate a transform that differs from the one the app performs. That module is covered by automated tests which check every species in every mode stays within a displayable colour range.

What this cannot do

These limits are not caveats added for safety; they are properties of the medium, and we would rather state them than let an impressive image imply more than it should.

• Ultraviolet, infrared and polarization cannot be shown. A phone screen has three emitters — red, green and blue — and a phone camera filters ultraviolet and infrared out before the sensor ever sees them. Where a species detects those, we map the invisible band onto a visible colour so its behaviour can be understood. That is a visualisation, not a reproduction, and each affected species says so on its own page.
• No simulation reproduces subjective experience. We can reproduce what information reaches the retina. What a brain then makes of it is not measurable by us or by anyone else.
• Adaptation and eye movement are not modelled. Real vision constantly adjusts to light and scans a scene. The filter shows a single steady state.
• Individuals vary. Published figures are population means from small samples. A specific animal may sit well away from them.

How this site is written and translated

We would rather describe this accurately than let the scale of the site imply more hand work than it received.

• The scientific content — every value, every citation — is written and reviewed by hand from the primary literature. It is not generated and not scraped.
• The editorial prose exists in about 70 languages. Seven of them (English, Spanish, French, German, Italian, Portuguese and Chinese) are authored directly. The rest are produced with machine assistance and then reviewed and corrected rather than published as-is.
• Species names are checked against native-language sources before publication, one language at a time, and the source for each name is recorded with its URL and date. A name that only exists as an English loanword is only used when a native source attests that loan. This matters more than it sounds: an automatic translation will confidently produce a name for a species that has none, or the name of a different animal entirely — a fault we have found and corrected in this corpus.
• Automated checks run over every built page before release: title and description presence and length, untranslated placeholders, correct text direction and language tags, valid structured data, per-species citations, complete hreflang groups, and duplicate titles or descriptions within a language. Separate checks cover grammar traps specific to individual languages, such as articles and sentence case. Any finding fails the release rather than being noticed later.
• A published page that turns out to be wrong is corrected at the source and rebuilt. Corrections are kept in the codebase so a later regeneration cannot silently undo them.

If you find an error — a mistranslated name, a misread figure, a source we have misrepresented — please write to [email protected]. Corrections are welcome and we act on them.

Who publishes this

Animal Vision is an independent project, not affiliated with any university, laboratory or journal. It is published by the developer of the Animal Vision mobile app, and it is funded by that app: a one-time purchase, optional per-species unlocks, and rewarded advertising. The research this site reports on is not ours — we cite it; we did not conduct it, and we have no financial relationship with any of the authors listed below. Where our reading of a paper is a simplification, the species page says so.

Content review: Marc Guirado Medonza.

See also: Guides · Support · Privacy Policy · Terms of Use

Sources by species

117 distinct works, 124 citations. Each also appears on that species' own article, in every language.

© 2026 Animal Vision · animalvision.app

Álvaro · Animal Ears · Lure Depth · Newborn Vision