CarbCam Changelog: all updates

All important CarbCam updates at a glance: new features, redesigns, new languages and improvements, grouped by month.

As of version 5.1.894, released September 29, 2026. Jump to: September 2026 · August 2026 · July 2026 · June 2026 · May 2026 · April 2026: native app · October 2025: project start · Backstory

• Hints by voice: hints for the AI can now be dictated by microphone. A recording window shows live what is recognized, a large Done button ends the recording.
• Permanent hints: frequent hints for the AI can be saved as "always send", then they go with every scan automatically.
• Faster packaged-product check: the nutrition cross-check for packaged products is now faster.
• Voice input for corrections too: the microphone is now also in the correction field after a scan.
• More stable start: a rare startup crash on some iPhones is fixed.
• Analysis in your language: the AI description, ingredients and explanations now come in your app language, read-aloud included.
• Nearby recommendations: the last card no longer slips under the navigation bar.
• Translations improved: a few remaining texts are now correctly translated.
• Nightscout API v3: now selectable, sign in with an access token.
• Google Gemini keys: new keys (AQ. format) now work.
• AI key: stray spaces or line breaks pasted in no longer break it.
• Nightscout: detects your own domain in the 10be field and offers a switch to self-hosted.
• OpenRouter: current models (Claude, GPT, Gemini) selectable directly, your own model stays saved.
• Feedback with photo: you can now attach a photo to a reply and optionally send a debug log; support photos appear right in the thread.
• OpenRouter model choice: your selected model is now actually used for the analysis, no longer the default model.
• Info popup: now closes reliably even with long text, the X is always reachable.
• AI models updated: new Claude, Gemini and GPT versions are now available.
• Translations: more incorrect texts in several languages fixed.

• More accurate values for packaged products: the values from the nutrition label are now applied exactly and are no longer overwritten by an automatic correction.
• Deleting without Nightscout: entries can now be reliably deleted in local-only mode too.
• History photos: locally saved photos load reliably again.
• Bolus reminder: detects boluses you have given more reliably, fewer reminders at the wrong time.
• Faster, more stable scan: the analysis starts across several parallel processes, even on a weak connection.
• Daily nutrition in the history: tapping the daily carb total shows kcal, fat, protein and FPE in a popup.
• Full-day analysis: long-pressing the daily total starts the AI evaluation of the whole day.
• Subscription start: usage indicators stay neutral during an active subscription (no premature yellow/red).
• “What's new”: can be hidden permanently and re-enabled in the settings.
• Mark a reference object: tap a photo to open it large and mark a scale object (e.g. dice, a coin or an AR marker) directly on the image. The AI uses it as a size reference and does not count it as carbohydrates. If the reference object is edible, you can include it via a checkbox.
• More accurate nutrition table: for packaged products the AI now picks up the legible table values better, reliably tells "per 100 g" apart from "per portion" and calculates the carbohydrates more accurately.
• Tip suggestions: tips you once entered for the AI (e.g. "celiac disease") are saved and offered as tappable suggestions on your next scan.
• Faster scan start: optimized so the results are ready even faster.
• Net quantity on packaged products: if the nutrition table and the printed net quantity are readable in the photo, the AI now uses exactly that amount for the portion size instead of an estimated standard portion.
• Plausibility warning: if a new estimate deviates strongly from the average of your similar meals, the app warns you before saving.
• Under the hood: the AI prompt is created server-side, so improvements to the analysis quality take effect immediately for all app versions.
• 34 languages: the app is now available in 34 languages, recent additions include Italian, Portuguese, Russian, Japanese, Turkish, Chinese, Korean, Arabic, Hindi, Thai and Vietnamese.
• Scan grouping: several scans can be combined into a single meal in the history, and statistics count the group as one meal.
• iOS redesign completed: scan result, history, statistics and subscription page now share the same card look as Android.
• Subscription & credits: usage and subscription status shown as a clear table, tapping the status card takes you straight to subscription & credits.
• Carb warning threshold: configurable warning before carbohydrates are forwarded to AAPS or Loop.
• Language picker: sorted alphabetically, switching languages now takes effect immediately.
• History: new time range filters for 6 months and 1 year.
• Optimizations and improvements: refined in many places and your feedback incorporated.

• Major design update: new card design throughout the app, big total-carbs display in the scan result, cleaned-up dialogs and automatic scrolling to the result.
• Invite friends: personal invitation code in the settings, both sides receive bonus scans.
• Nightscout: carbohydrates can be written to Nightscout as a carb correction, even without a loop system.
• Carb warning threshold: the daily goal warning now appears directly in the result view.
• Feedback with photo: photos or screenshots can be attached to feedback.
• AAPS, Loop, iAPS and Trio: step-by-step guides linked directly from the settings.
• New language: Czech.
• Settings: AI model and personal API key are saved immediately, local photos are kept permanently.

• Backup: export and import of all data and settings on Android and iOS.
• Restaurant reminder: the app reminds you to scan when you are at a restaurant, with a home zone: the reminder stays silent at home, and the home area can be set on a map with an adjustable radius.
• Cooked or raw: for pasta, rice and similar foods the app asks instead of silently overwriting values; the same applies to packaged products (label or database).
• Barcode scan extended: now also reads fat, protein and calories.
• Faster analyses: photos are uploaded in the background while you are still taking them.
• OpenRouter: additional AI provider with your own API key (BYOK).
• Statistics: the day filter now works with calendar days instead of a rolling 24 hours.
• Clearer error messages: specific hints for connection and configuration problems instead of raw provider text.

• Statistics as its own tab: zoomable charts, time-of-day evaluation, carb/kcal/FPU bars per day, pie chart and top foods.
• Fat and protein everywhere: carried through the whole pipeline, including forwarding to AAPS, xDrip+, Trio and iAPS.
• Daily goals: personal targets for carbs, calories and more, with a warning when you exceed them.
• Subscription inside the app: six subscription tiers and extra credit packs via App Store and Google Play, Stripe for purchases in the browser.
• 10be hosting bonus: now counts in addition to your subscription instead of replacing it.
• Version 5.0: the app reads your Nightscout profile (ICR, ISF, basal), rates meals, learns a personalized carb correction and offers the doctor mode with pattern detection for the next doctor visit.
• Insights: new evaluation page in the statistics plus glycemic metrics with a selectable time range (7, 14 or 30 days).
• Privacy: analytics and crash reports only with explicit consent, off by default.

• Complete rewrite: CarbCam is rebuilt as a native app: Android with Jetpack Compose, iOS with SwiftUI, shared logic via Kotlin Multiplatform. From mid-April the native app fully replaces the first React version.
• iOS version: CarbCam arrives on the iPhone for the first time, beta via TestFlight.
• 7 languages: German, English, Spanish, French, Polish, Dutch and Hebrew.
• Barcode scanner: capture packaged products via barcode, with a database cascade (Open Food Facts, USDA, AI).
• First subscription: analysis quota with purchase via Stripe in the browser.
• Feedback threads: feedback straight from the app, with replies from the team and a history.
• Nearby sharing: share meals with people close to you, with details, photo and a report function.
• Meal reminder: notification at meal time that takes you straight into the app.

• CarbCam launch: the first versions run as a React app (Capacitor) on Android: AI-powered photo analysis of meals with estimates for carbohydrates, calories and portion sizes, up to 2 photos per meal.
• Nightscout from day one: meal upload as a note to your own Nightscout instance and BG curve for each meal.
• BG analysis and comparison: view the blood glucose curve after eating and compare meals with each other.
• FPU calculation: fat-protein units with a delay hint for meals rich in fat and protein.
• The name CarbCam: rebranding with the new fork-and-spoon logo in the 10be style.
• Photo upload and cache: meal photos are stored on the ns.10be.de server, repeated analyses are served from the cache.
• Full-text search in the meal history and feedback straight from the app.

• 2018: the EU research project GoCarb shows that estimating carbs from a photo is possible, but it is never released as an app for everyone.
• 2021: the idea keeps coming back: new approaches and sketches for an app of our own, again and again.
• 2024: first versions using AI are built but put on hold: expanding 10be takes priority.
• 2025: work on the app picks up speed again.
• Spring 2026: after the big ns.10be.de overhaul, every free minute goes into the CarbCam app: in April, development starts in today's repository.