Vision Assistant AI

# What's New in This Version ## 1. Describe now covers the whole scene, not just what's closest Double-tap Describe used to focus mostly on the nearest object. It now describes **both**: - **Foreground** — obstacles, objects within reach, anything underfoot - **Background** — the wider room, doorways, or people further away So a single description gives you the full picture of your surroundings, not just the closest item. ## 2. Find now searches the whole camera view, foreground and background Previously, Find used a zoomed-in camera view that could crop out anything not centered and close. Find now uses the full, wide field of view, so it can locate your object whether it's close by or further into the background. ## 3. Find distance is now actually accurate Distance used to say "about 3 feet" for almost everything, regardless of how far the object really was. This was a bug in how distance was estimated — it's now calculated properly, per object and per frame, using the object's real size on screen. You should now get distance estimates that actually reflect how far the object is (closer objects read as closer, farther objects read as farther). *Note: if an object can't be precisely measured (e.g. Find falls back to a lower-detail detection mode), the app will now say something like "found it nearby" instead of guessing a specific, potentially wrong number.* ## 4. Find direction (left/right) is fixed There was a bug where Find could occasionally announce the wrong side — e.g. saying an object is on your right when it's actually on your left. This was caused by an internal orientation mismatch and has been corrected. Directional guidance should now reliably match what's actually in front of you. ## 5. Describe is now conversational — ask follow-up questions After Describe finishes speaking, the app will now prompt: *"You can ask a question about what I described, or say done to finish."* You can then ask multiple follow-up questions about that same scene (e.g. "what color is it?", "how many people are there?", "is there a chair nearby?") and get relevant spoken answers, without needing to re-describe the whole scene each time. - Say **"done"**, **"stop"**, or similar to end the conversation whenever you're ready. - Starting a new Describe, Find, or Read action will automatically end the current follow-up conversation.