fud-ai

Introduction: Fud AI is a Free & Open Sourced AI Calorie Tracker for iOS & Android

Tags:

Fud AI Logo

Fud AI

Eat Smart, Live Better
Snap, speak, or type your food — AI handles the rest.

Open-source, privacy-first calorie tracker for iOS and Android. Bring your own AI provider — 13 supported including Gemini, OpenAI, Claude, Grok, Groq, Hugging Face, Fireworks AI, DeepInfra, Mistral, and any custom OpenAI-compatible endpoint — or use optional Fud AI Premium on iOS for hosted Gemini + Deepgram access. Snap a meal, share a food photo into Fud AI, scan a barcode, combine two camera shots, add a note to a camera or library photo, ask your AI coach how to hit your goal, speak your lunch, or use Siri Shortcuts on iOS to log food and weight. On supported iPhones, food-description analysis for text, voice-transcribed, and Siri food logs can use Apple Intelligence on-device as the final fallback after Premium Gemini or BYOK provider/fallback attempts fail. No accounts, no cloud sync, no tracking.

iOS 4.3 build 24 keeps the Apple Intelligence fallback release and adds Polish localization fixes for dynamic nutrient names, meal types, Settings labels, "Log" wording, and locale-aware date/time display.

Android 2.2.2 build 25 fixes text food entry keyboard focus, adds native speech fallback when Android on-device language support is unavailable, and recommends Groq (Whisper) or Deepgram for more consistent voice transcription across Android phones.

App Store · Google Play · Website · Report an Issue · Request a Feature

Logging

Snap food — camera identifies meals and estimates nutrition
Camera + Note — add a description with the photo for better accuracy
Camera + Camera — capture two images for one analysis, useful for front/back packaging or food + label combos
iOS Share Extension — send a food photo from Photos or another app directly into Fud AI for review and logging
Nutrition label scan — reads packaging for precise per-serving data
Barcode lookup — scan packaged foods on iOS and Android and fill nutrition from Open Food Facts when product data is available
Photo library — analyze existing photos
Photo library + Note — pick an existing photo and add context before AI analysis
Text input — type food descriptions
Voice input — speak your meals hands-free (6 STT options with per-provider language selection, see below)
iOS Siri Shortcuts — say phrases like "Log food in Fud AI", "Calories today in Fud AI", or "Log my weight in Fud AI"; the phrase guide lives in Settings → Siri Phrases
Manual Entry — log known calories and macros without AI
Smart serving units — AI can show slices, pieces, cups, ml, or other visible serving units while grams stay the source of truth
Review nutrition unlock — correct calories, macros, and detailed nutrients before logging, then lock again so serving changes scale from your edits
Meal What if? — preview how a reviewed meal changes today's calories and macros, then ask AI for a practical suggestion before logging
Saved Meals — Recents, Frequent, and Favorites with safer swipe actions, search, and drag-to-reorder

Intelligence

AI Coach tab — multi-turn chat with memory. Coach sees your profile, weight history, food log, today's date/timezone, and richer meal details, then answers questions like "what's my expected weight in 30 days?" or "how do I lose 2 kg?". Coach also supports camera/photo attachments on Android. Memory persists across launches; Reset button starts a fresh conversation. Long-press any reply to copy.
AI Access — choose free Bring Your Own Key or optional Fud AI Premium on iOS. BYOK lets you pick provider, model, fallback, custom instructions, and speech language directly on device; Premium uses Fud AI's Gemini + Deepgram proxy with daily safety limits and hides BYOK-only controls until you switch back.
Apple Intelligence fallback — on supported iPhones, food-description analysis for text, voice-transcribed, and Siri food logs can use Apple Intelligence on-device as the final fallback after Premium Gemini or BYOK provider/fallback attempts fail.
AI optional nutrient goals — estimate detailed nutrient goals from profile data without changing calorie/protein/carbs/fat formulas.
Goal-aware prompt chips — suggested questions change based on whether your goal is Lose / Gain / Maintain
Thermodynamic weight forecast — expected weight at 30/60/90 days, predicted vs observed weekly change, days-to-goal, under-logging detection. Surfaced through Coach as live context on every turn.
Resilient requests — transient provider overloads (503 / 529 / 429) auto-retry with 1s / 2s / 4s exponential backoff across both food analysis and Coach chat, so short spikes resolve invisibly

Tracking

Expanded nutrients per entry — macros plus sugar, fiber, fats, cholesterol, sodium, potassium, calcium, iron, magnesium, zinc, vitamins, folate, omega-3, and more when available
Custom Home nutrient cards — swap the top cards from protein/carbs/fat to fiber, sodium, vitamin D, calcium, or other tracked nutrients
Optional nutrient goals — set or AI-estimate goals for the non-macro nutrients; these stay separate from the calorie and macro calculator
Scrollable week calendar — swipe to any past week, configurable start day
Food log sorting — keep the default grouped view, or sort meal sections by latest logging order from the Home screen
Progress charts — weight trends, calorie history, macro averages (1W to All Time)
Progress summaries — weight and body-fat ranges show average and net change for the selected week, month, or longer window
Decimal nutrition totals — macros and detailed nutrients preserve decimal precision in logs, Home, widgets, and View More
Weight History — tap-to-delete past entries, syncs deletion to Apple Health
Goal tracking — set target weight, BMR/TDEE auto-calculation; goal-reached alert fires from both manual logs and Apple Health reads
Adaptive Goals (Experimental) — optional weekly calorie correction from observed weight trend; pinned macros stay pinned and unlocked macros auto-balance

Health & platform

Apple Health — bidirectional sync for body measurements + nutrition types written per meal; Siri food/weight logs write through the same HealthKit paths, and Experimental Energy Burn Goals can estimate calorie targets from active/total energy while macros stay editable
Health Connect — Android sync for nutrition, weight, and body fat, with permission reconciliation and backfill support; Experimental Energy Burn Goals can use recent energy data for calorie targets
Apple Watch — watchOS app and complications show calories and macros at a glance
Widgets — iOS Home Screen has Fud AI in Small, Medium, and Large plus a small-only Fud AI Protein widget; Lock Screen widgets stay separate, and Android Glance widgets update when you log
Android 2.2.2 update — Text food entry keeps the keyboard open while example placeholders rotate; native Android voice logging falls back when on-device language support is unavailable
Share the App — native iOS share sheet from About → forwards App Store URL plus a personalized message and fud-ai.app link; message body localized into all 16 iOS languages
Update check — About shows the installed app version, opens the App Store / Play Store when a newer version is available, and shows a tab dot for pending updates
Theme color — iOS and Android Settings let users change the app accent, with matching home screen / launcher icons
Languages — iOS supports 16 languages: Arabic, Azerbaijani, Dutch, English, French, German, Hindi, Italian, Japanese, Korean, Polish, Portuguese (Brazil), Romanian, Russian, Simplified Chinese, Spanish. Android supports the same set except Polish. The app auto-selects by the phone's Language setting; iOS 4.3 build 24 improves Polish dynamic labels and locale date/time formatting.
Meal reminders — customizable breakfast, lunch, dinner notifications
Dark mode — system, light, or dark
Metric & imperial units

AI Providers

Pick any of the 13 LLM providers for food analysis, meal what-if suggestions, optional nutrient-goal estimation, and Coach chat. Free Gemini keys are available at aistudio.google.com/apikey. In BYOK mode, requests go directly from your device to the provider you configure. On iOS, optional Fud AI Premium can instead use Fud AI's hosted Gemini + Deepgram proxy with daily safety limits. For text, voice-transcribed, and Siri food descriptions on supported iPhones, Apple Intelligence can run on-device only as the last fallback after Premium Gemini or BYOK provider/fallback attempts fail.

Provider	Format	Highlight	Needs API Key
Google Gemini	Gemini API	Gemini 3.5 Flash / 2.5 Pro / Flash	Yes
OpenAI	OpenAI	GPT-5 / 4o	Yes
Anthropic Claude	Messages API	Sonnet 4.6 (default) / Opus 4.7 / Haiku 4.5	Yes
xAI Grok	OpenAI-compatible	Grok 4	Yes
OpenRouter	OpenAI-compatible	Any model, free-form IDs	Yes
Together AI	OpenAI-compatible	Llama 4, Qwen VL	Yes
Groq	OpenAI-compatible	Llama 4 Scout, very fast	Yes
Hugging Face	OpenAI-compatible	Gemma 3, Qwen VL, Llama Vision (open-weight router, free-form IDs)	Yes
Fireworks AI	OpenAI-compatible	Qwen VL, Llama Vision, Phi-3 Vision	Yes
DeepInfra	OpenAI-compatible	Gemma 3, Llama Vision, Qwen VL	Yes
Mistral	OpenAI-compatible	Pixtral Large / 12B (open-weight)	Yes
Ollama	OpenAI-compatible (local)	Llama 3.2 Vision, LLaVA, Moondream — runs on your Mac	No
Custom (OpenAI-compatible)	OpenAI-compatible	You set base URL + free-form model name	Optional

Speech-to-Text Providers

Pick how voice input is transcribed. Native iOS / Android is the default — free, on-device where supported, real-time. On Android, native speech first tries the on-device language path, then falls back to Android recognition with network/provider defaults if the phone lacks offline support for that language. Each provider has its own language setting: use Provider Auto, Use Device Language, or an explicit language hint where supported.

Provider	Notes
Native iOS / Android (On-Device)	Free, offline where the phone supports the selected language, real-time partial results
Gemini Audio	Batch audio transcription through Gemini for BYOK users
OpenAI Whisper	Whisper-1 via `/v1/audio/transcriptions`
Groq (Whisper)	Whisper-large-v3, very fast, has a free tier
Deepgram	Nova-3, fast and accurate
AssemblyAI	Universal model, strong accuracy, free tier

For Android phones where native speech is inconsistent, Groq (Whisper) or Deepgram are recommended alternatives; the developer currently uses Groq.

API keys are stored encrypted on-device: iOS Keychain on iOS and EncryptedSharedPreferences backed by Android Keystore on Android.

How It Works

Photo / Text / Voice
        │
        ▼
  BYOK provider API or Premium proxy
        │
        ├── BYOK provider fallback if configured
        └── iOS Apple Intelligence final fallback for text / voice transcript / Siri food descriptions
        │
        ▼
  JSON nutrition response
        │
        ▼
  User reviews & edits
        │
        ▼
  FoodStore.addEntry()  ──▶  UserDefaults (local) + Apple Health (optional)

For the Coach chat, every turn builds a slim system prompt from your live profile, BMR formula in use, computed forecast, today's date/timezone, and a one-line snapshot of available data. Coach then pulls any date range of weight, body fat, calorie totals, or food entries on demand via tool calling — ask "what was my weight in March?" or "show me my body fat trend over the last 6 months" and it fetches exactly the slice it needs, including meal source, meal type, serving size, and micronutrients.

Screenshots

An eight-screen walkthrough of the current app flow — from opening the dashboard to detailed nutrition and progress.

01 · Home · Dashboard _{Daily calorie ring, selected Home nutrient cards, and today's logged meals grouped by meal type. Week strip at the top for date navigation.}	02 · Log · Options _{Tap + to open Camera, Camera + Note, Camera + Camera, Nutrition Label, Barcode, Photos, Text, Voice, Manual Entry, Saved Meals, or Copy from Day.}	03 · Snap · Capture _{Point and shoot. The image is sent to your chosen AI provider; nutrition estimates come back within a few seconds.}
04 · Review · Edit _{Review the AI's guess, unlock nutrition if values need correction, adjust the serving size (everything recalculates live), preview "What if?" impact, and pick a meal type before logging.}	05 · Meals · Log _{The day's entries grouped by breakfast / lunch / dinner / snack. Swipe to delete, tap to edit any entry.}	06 · Coach · AI Chat _{Multi-turn conversation with full context of your profile, weight history, food log, and forecast. Ask "what should I eat?" or "expected weight in 30 days?".}
07 · Progress · Charts _{Weight trend with goal line, calorie history (intake vs. goal), and macro averages. Time ranges span 1 week to all time.}	08 · View More · Nutrients _{Detailed nutrition shows macro totals plus optional nutrients such as fiber, sugar, sodium, minerals, vitamins, and fats.}

Calorie & Macro Calculation

The app calculates personalized daily targets using established nutrition science formulas:

Step	Formula	Details
BMR	Katch-McArdle	`370 + 21.6 × lean mass (kg)` — used when body fat % is known
BMR	Mifflin-St Jeor	`10w + 6.25h − 5a ± 5` — fallback when body fat is unknown
TDEE	BMR × activity	Multiplier ranges from 1.2 (sedentary) to 2.0 (extra active)
Daily Calories	TDEE + adjustment	Adjustment = `weeklyChangeKg × 7700 / 7` (deficit or surplus)
Protein	Activity + goal	`0.8 – 2.2 g/kg` body weight by activity, plus +0.2 g/kg during cutting phase (Helms et al 2014); when body fat % is known, Activity Level also shows the equivalent g/kg lean-mass multiplier
Fat	Fixed ratio	`0.6 g/kg` body weight
Carbs	Auto-balanced	Remainder from calories − protein − fat (any macro can be pinned; max 2 pinned)

All values can be manually overridden in Settings, with a Recalculate Goals button to snap back to formula defaults.

Architecture

Component	Details
Language	Swift 5, SwiftUI, iOS 17.6+
Storage	UserDefaults (local JSON), Keychain (API keys)
AI	`GeminiService` for food + label analysis, `ChatService` for multi-turn Coach chat, both route across all 13 providers
Speech	Native `SFSpeechRecognizer` / Android `SpeechRecognizer` or remote providers via `SpeechService` (m4a upload)
Health	HealthKit / Health Connect read-write paths for body measurements and meal nutrition, with UUID-tagged samples for safe delete
Pattern	`@Observable` + `.environment()`, main actor isolation
Localization	`Localizable.xcstrings` (String Catalog), 16 iOS languages, auto-selected by iPhone's system language
Dependencies	Native platform frameworks; app data and API keys remain local

Repo Layout

fud-ai/
├── ios/          # SwiftUI iOS app (shipping on App Store, v4.3)
├── android/      # Kotlin + Jetpack Compose app (min SDK 26 / Android 8.0, v2.2.2)
├── web/          # Marketing site — https://fud-ai.app (static HTML/CSS, Vercel)
├── APPSTORE.md   # App Store Connect listing copy (iOS)
├── PLAYSTORE.md  # Google Play Console listing copy (Android)
└── README, LICENSE, CONTRIBUTING, SECURITY, CLAUDE.md, .github/

Source Layout (iOS)

ios/
├── calorietracker.xcodeproj/         # Xcode project
├── calorietrackerTests/              # Unit test target (boilerplate)
├── calorietrackerUITests/            # UI test target (boilerplate)
├── FudAIWidgets/                     # Widget extension target (Home + Lock Screen)
├── screenshots/                      # App Store screenshot sources
└── calorietracker/
    ├── calorietrackerApp.swift       # Entry point, environment setup
    ├── ContentView.swift             # 5-tab layout (Home, Progress, Coach, Settings, About)
    ├── Localizable.xcstrings         # String Catalog, 16 languages
    ├── Models/
    │   ├── AIProvider.swift          # 13 LLM providers, model lists, settings
    │   ├── SpeechProvider.swift      # 6 STT options + Keychain settings
    │   ├── ChatMessage.swift         # Coach chat message model
    │   ├── UserProfile.swift         # BMR/TDEE/macro calculations
    │   ├── FoodEntry.swift           # Food item with macros and expanded optional nutrients
    │   └── WeightEntry.swift         # Weight log entry
    ├── Views/
    │   ├── OnboardingView.swift      # 15-step onboarding flow
    │   ├── ChatView.swift            # Coach tab: bubbles, prompt chips, reset
    │   ├── FoodResultView.swift      # AI result review & edit
    │   ├── RecentsView.swift         # Saved Meals (Recents / Frequent / Favorites)
    │   ├── VoiceInputView.swift      # Native + remote STT routing
    │   ├── HomeComponents.swift      # Week strip, macro cards
    │   └── ProgressComponents.swift  # Charts, weight history
    ├── Services/
    │   ├── GeminiService.swift       # Food/label analysis, routes 13 providers
    │   ├── ChatService.swift         # Multi-turn Coach chat, routes 13 providers
    │   ├── SpeechService.swift       # Remote STT router (Gemini / OpenAI / Groq / Deepgram / AssemblyAI)
    │   ├── WeightAnalysisService.swift # Thermodynamic weight-forecast math
    │   ├── KeychainHelper.swift      # iOS Keychain wrapper
    └── Stores/
        ├── FoodStore.swift            # Food CRUD + favorites
        ├── WeightStore.swift          # Weight CRUD (auto-syncs profile weight)
        ├── ProfileStore.swift         # @Observable wrapper over UserProfile
        ├── ChatStore.swift            # Coach chat history (persisted locally)
        ├── NotificationManager.swift  # Notification scheduler
        └── HealthKitManager.swift     # Apple Health bridge (body + nutrition)

Build & Run

# Clone
git clone https://github.com/apoorvdarshan/fud-ai.git
cd fud-ai

iOS

xcodebuild -project ios/calorietracker.xcodeproj \
  -scheme calorietracker \
  -destination 'platform=iOS Simulator,name=iPhone 17 Pro' build

Open ios/calorietracker.xcodeproj in Xcode, select your device, and run.

Android

Open android/ in Android Studio (Narwhal or newer), let Gradle sync, hit ▶ Run. Or from the CLI:

export JAVA_HOME="/Applications/Android Studio.app/Contents/jbr/Contents/Home"
cd android
./gradlew :app:assembleDebug
adb install -r app/build/outputs/apk/debug/app-debug.apk
adb shell am start -n com.apoorvdarshan.calorietracker/.MainActivity

First launch walks you through onboarding (gender, birthday, height/weight with metric/imperial toggle, body fat %, activity with protein target preview, goal, goal speed, notifications, Apple Health / Health Connect, AI access setup, review). A free Gemini key is available at aistudio.google.com/apikey, or iOS users can choose optional Fud AI Premium. You can switch anytime in Settings → AI Access; BYOK provider/fallback/speech settings appear only while BYOK is selected.

Contributing

See CONTRIBUTING.md for guidelines. Bug reports and feature requests welcome.

Adding a new translation? Open ios/calorietracker/Localizable.xcstrings in Xcode and fill in your language column — everything else is already wired.

Security

See SECURITY.md. Use private vulnerability reporting for sensitive issues.

Privacy

No accounts, no cloud sync, no analytics. BYOK API keys are encrypted on-device and requests go directly to the provider you choose; optional iOS Premium requests go through Fud AI's Gemini + Deepgram proxy only for the request being processed. On supported iPhones, text, voice-transcribed, and Siri food descriptions can be processed by Apple Intelligence on-device as the final fallback after Premium Gemini or BYOK provider/fallback attempts fail. Barcode lookup sends the scanned barcode to Open Food Facts and stores the returned nutrition locally. Optional nutrient goals, Adaptive Goals preferences, Home nutrient-card choices, saved review nutrition edits, food photos from the app or iOS Share Extension, iOS Siri/App Intent food or weight phrases, cached thumbnails, widget snapshots, and Apple Watch nutrition snapshots are local preferences/data except for the specific provider, Premium proxy, or final on-device Apple Intelligence request needed to analyze a food description. AI estimation sends only the context needed for that request, such as reviewed meal + daily-total context for a meal what-if suggestion or Siri food text for Siri food logging. Apple Health / Health Connect energy-burn goals read active/total energy only after the user enables that setting. Delete All Data is local-only — it wipes the app's storage (food log, weight log, body-fat log, profile, Coach chat, saved meals, API keys, widget / Watch snapshot) but never touches Apple Health or Health Connect. Samples you've synced are yours; if you want them cleaned up, do it from Health / Health Connect settings. See Privacy Policy.