If you have ever tried to track macros through a full training block, you know the drill: open the app, search for “Chipotle chicken bowl,” scroll past forty almost-right entries, pick one, adjust the grams, repeat four times a day. It is not hard. It is just tedious enough that people stop doing it by week three.
TrakMac takes a different approach. Tap the mic. Say what you ate. The app estimates the macros using an AI model, shows you the numbers, and you confirm. That is the whole loop.
Why speaking works better than scanning
Barcode scanning is great if every meal comes in a package. It falls apart the moment you eat something restaurant-prepared, homemade, or assembled on a plate. Which, for anyone eating real food around serious training, is every meal.
Voice input solves three problems at once:
- Speed. Ten seconds of talking beats two minutes of database scrolling.
- Generality. A description like “grilled chicken, maybe six ounces, cup of jasmine rice, broccoli with olive oil” covers almost anything you can plate. No database entry needed.
- Accuracy where it counts. For tracking-to-optimize-training (not medical nutrition), ballpark-right numbers you actually log are more useful than perfect numbers you skip.
But is AI accurate enough?
For fitness tracking, yes. TrakMac uses Claude to estimate macros from your description, trained on food composition data and typical portion sizes. For common meals — chicken breast, rice, oatmeal, a Chipotle bowl — estimates come in within ~10% of a weighed log.
For edge cases (a dish you made up last weekend, an unusual cut, a huge restaurant plate), the app tells you confidence is low and lets you adjust. That’s the point: voice-first doesn’t mean blind. It means you get a smart first draft instead of starting from a search bar.
If you need exact numbers for medical reasons, weigh your food and log it that way. If you’re tracking to hit protein and stay under a calorie budget while you build strength — which is what most lifters actually need — close-enough estimates produced in ten seconds beat exact numbers produced never.
How TrakMac is different from MyFitnessPal
MyFitnessPal is a food-database search engine. TrakMac is a voice interface plus a body-type-aware target calculator. Three specific differences:
- Voice-first logging. No search bar. Describe the meal, confirm, done.
- Body-type-aware targets. Targets aren’t just Mifflin-St Jeor plus activity. TrakMac asks what you actually train — bench max, mile time, muscle-ups, echo bike — and tunes protein and calorie baselines to match.
- Not trying to be a clinical tool. TrakMac is for people who lift, want reasonable numbers, and want to get on with their day. It is explicit about being “close-enough” — which for training, is what you actually need.
Get on the list
Join the beta. iOS first, public beta opening this year.