EchoSight
A phone-based visual assistant for blind and low-vision users that asks the same question twice to confirm what it actually saw — so it tells you what's on the can, not what it guessed was on the can.
Blind and low-vision users who rely on their phone to read labels, signs, prescriptions, and describe unfamiliar scenes
- Snap or upload any photo to get a focused description that confirms it read the relevant text twice before answering
- Targeted follow-up Q&A for the specific details that matter ('is this medication the 10mg or 20mg?')
- Document and label mode tuned for expiry dates, dosage text, and small print that mainstream apps miss
- Optional share-to-caregiver mode that texts the verified description to a family member when needed
VLMs routinely misread text and small visual details — the prompt-echoing finding shows up to 17.5 accuracy points of recoverable error, exactly where assistive tech users get burned by confidently wrong answers
Validated by a mature ecosystem of AI visual assistants (Be My Eyes with GPT-4o, Microsoft Seeing AI, Google Lookout, Envision, SightSense, Ally, Aira) targeting an audience of ~1.3B people with disabilities globally.10 Breakthrough AI Tools Empowering Blind And Low-Vision Users ↗
Market is crowded with free offerings from Microsoft (Seeing AI) and Google (Lookout), plus Be My Eyes, Envision, Aira, SightSense, and Ally all doing scene/text description; 'prompt echoing' is a UX refinement, not a defensible moat.Top 5 Apps for Visual Impairment in 2026 - ATandMe ↗Envision - Assistive Technology for Blind and Low Vision ↗
Direct competitors Seeing AI and Google Lookout are completely free, and Be My Eyes/Access AI have introduced free AI tiers; monetizing a phone-only VLM wrapper against subsidized incumbents is very difficult, and the blind/low-vision demographic has limited disposable income.Top 10 AI Tools for the Blind and the Visually Impaired ↗9 Best AI Tools for Blind Users in 2026 (Free + Paid, Tested) ↗
Vision loss is a permanent need and VLM hallucination/misread errors on labels and small text will persist as models scale; accessibility mandates and aging populations provide durable tailwinds.Lookout - Assisted vision - Apps on Google Play ↗
Buildable as a thin mobile wrapper around existing VLM APIs with double-pass inference, but requires careful accessibility testing, screen-reader integration, and latency optimization to be usable in real-world pointing scenarios.Lookout - Assisted vision - Apps on Google Play ↗SightSense ↗