Original Summary

I wanted to lose weight and started logging my calories in Google Sheets, but over time the process felt too clunky and time-consuming. I wanted something where I could type something like “232 grams of pasta, 360 kcal per 100 grams” and get the numbers already calculated and added to a table. Like, open the app, type it in, close it. I looked at existing apps, but every fucking one makes you press a dozen buttons just to log a fucking apple. Cal AI even makes you fill out a 20–30 question survey during signup before letting you try the app. Anyway, this was a very specific pain point for me. I hate apps where you have to wrap your head around the interface. Looking at you, MyFitnessPal. So I built my own app. Even came up with a funny name, or so I thought at the time. Initially, I had no plans to monetize it in any way. It was just for personal use. Half a year has passed since then, I’ve added a few more features, kept using it myself, and actually lost about 10 kg, just like I wanted. At some point, I decided to release the app publicly. Maybe someone else will like it too. link: https://onlyfats.fit The main thing is very simple: you type something in natural language, like “one green apple,” “three chicken eggs,” or “232 grams of pasta, 360 kcal per 100 grams,” and get an entry in a table. I wanted to pick the fastest model possible, because the whole idea is to spend as little time as possible on the logging itself. It would be stupid to build a fast interface and then wait 10 seconds for the model to respond. Gemini Flash Lite worked best for this. Though a few times, even that took 6–10 seconds to respond because it was overloaded, so I added a fallback to a few other providers. Then I added photo analysis. I hadn’t planned on it at all. I just already had text parsing and got curious about how well this would work. A problem with meals came up pretty quickly: the model is pretty bad at figuring out how much food there actually is from a photo. So I ended up having it calculate calories per 100 grams and letting the user enter the weight. But unexpectedly, the most useful application of this feature turned out to be something else entirely. My wife, who was the pickiest tester, started using photos to quickly copy calories and macros from food labels instead of analyzing meals. And that turned out to be really convenient. Especially when you scan a product’s barcode and it’s not in the database. Instead of entering everything manually, you just take a photo of the label. I also noticed that I use some products all the time. Sauces, for example. Having to type “such-and-such sauce, 39 grams, 350 kcal per 100 grams, protein/fat/carbs 1/30/22 per 100” every time got annoying pretty quickly too, so I added the option to save frequently used products. After that, you just need to type the weight and the first few letters of the name. Anyway, building this turned out to be a pretty interesting experience. Especially when you make something bec

中文概览

中文标题： 又一个 AI 卡路里计算器。

作者为简化卡路里记录而开发的应用：用自然语言输入如“232克意面，每100克360千卡”即可计算并记入表格；选用Gemini Flash Lite等模型，支持照片分析、营养标签读取和保存常用食物，已公开。

  • 情报分类：开源项目与落地
  • 分类依据：个人AI卡路里记录应用发布，属项目落地分享
  • 信息来源：Reddit · SideProject
  • 发布时间：2026/9/16 20:06:05