Apple's PICO Image Codec Ignites Debate Over 'Texture Synthesis' and Hallucination

Apple researchers have introduced PICO (Perceptual Image Codec), a learned image compression model optimized directly for the human visual system and on-device neural runtimes. PICO claims to achieve 2.3–3x bitrate savings against traditional standards like VVC, AV1, and JPEG-AI, while decoding a 12-megapixel image in just 150 milliseconds on an iPhone 17 Pro Max. However, the codec's reliance on generative reconstruction has triggered a deep philosophical and practical divide within the developer community.

To achieve ultra-low bitrates (sub-0.3 bits per pixel), PICO utilizes "texture synthesis" to reconstruct micro-details. Instead of preserving pixel-perfect data, the decoder identifies a textural quality and generates plausible details—such as individual clothing fibers, hair strands, or tree branches—from scratch. Critics argue this crosses the line from lossy compression into fabrication. As klodolph pointed out when reviewing the demo images: "In the PICO version, it looks just completely wrong to me. The yarn structure has been replaced with a bunch of fuzzy strips." User mrob warned of the deceptive nature of these artifacts: "Newer codecs create artifacts that can be mistaken for part of the original image. That's a heavy price to pay for improved compression efficiency."

This generative behavior raises serious concerns regarding deterministic decoding and forensic integrity. Unlike traditional JPEG compression, where artifacts (such as blockiness or ringing) are obviously artificial, PICO's synthesized details look highly realistic but are factually incorrect. Commenters noted that this could have major legal ramifications if images are used as evidence in court, or if a user zooms in to evaluate a product's texture for online shopping. As crazygringo writes: "It is essentially hallucination of details on a micro scale... we see blurriness as being more 'honest' ... whereas with textural hallucination, it is no longer clear what is being filled in versus what is original."

Practically, engineers also questioned PICO's 150ms decoding latency, noting that on modern mobile devices capturing 48MP or 50MP photos, linear scaling would result in a lag of over half a second per image, leading to battery drain and a sluggish user experience in photo gallery applications.