Welmi
Welmi is a calorie counter, an app in which a person records what they eat and the app adds up the calories and the protein, carbohydrate and fat against a daily target. The difference from the older apps in this category is the input: the user takes a photo of the plate and an AI model names the foods and estimates the amounts, or the user says the meal aloud. The app also has a GLP-1 diary. GLP-1 medicines such as semaglutide and tirzepatide are weekly injections prescribed for weight loss; the app records the dose, sends injection reminders and logs side effects, and its store page says that it does not recommend or calculate doses. The seller is T-SHAPED APPS LTD, a company with its registered office in Paphos, Cyprus, which also sells a period tracker and a gym workout planner.
The app launched in March 2025. Adapty, the subscription software company that records the app's purchases and that also runs its ad attribution, published two case studies. The first, dated 22 December 2025, gives a dated series of monthly recurring revenue (MRR, the subscription revenue of one month): US$60,000 in August 2025, US$150,000 in October 2025 and over US$200,000 in November 2025, eight months after launch. The second, dated 7 April 2026, quotes the co-founder, named in it only as Sergey, saying that revenue went "from around $30–50K to $300K in about four months" after the company began taking weekly advances on its App Store payouts instead of waiting 45 days, with no outside capital. A contact-data directory lists Sergei Tokarev as co-founder and CEO of Welmi.ai and counts 6 employees; no other downloaded source gives the full name or the headcount.
This page grades every number by its evidence. The headline figure is grade A because the company that processes the subscriptions stated it, as with the Wave AI story on this site. The headcount and the founder's full name rest on a directory and are marked as such. The page also shows five earlier health and nutrition startups from the Loot Drop database that failed with a similar promise.

The numbers, graded
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Monthly recurring revenue, November 2025 | over US$200,000 | A | Adapty's case study published on 22 December 2025, in its timeline: "November 2025: $200K+ MRR". Adapty's software records the app's subscriptions. The same page says the team went from its first organic sale to over US$200,000 in MRR within eight months of the March 2025 launch. | [1] |
| Run rate from the November 2025 figure | about US$2.4 million a year | calc | Calculated: 200,000 times 12. The headline figure on this page. | [1] |
| Monthly recurring revenue, August 2025 | US$60,000 | A | Same Adapty timeline: "August 2025: $60K MRR". | [1] |
| Monthly recurring revenue, October 2025 | US$150,000 | A | Same Adapty timeline: "October 2025: $150K MRR". | [1] |
| Growth from August to November 2025 | about 3.3 times in 3 months | calc | Calculated: 200,000 divided by 60,000 equals 3.33. Adapty's headline for the same period is "from $60K to $200K MRR in 3 months". | [1] |
| Revenue growth reported in April 2026 | from around US$30,000 to 50,000 to US$300,000 in about four months | A | Adapty's second case study, published 7 April 2026, quoting the co-founder: "We went from around $30–50K to $300K in about four months." The page calls it revenue and does not name the months or say per month in that sentence; its highlights say "Grew revenue from $30-50K to $300K in 3-4 months". The starting range is below the US$60,000 MRR that the first case study gives for August 2025, so the two pages do not describe the same months in a way this page can reconcile. | [2] |
| Outside funding | none | A | Adapty, April 2026: "They didn't raise outside capital or take on debt." The company used Adapty Finance, which advances up to 85% of unpaid store earnings for a fixed fee, with no equity. | [2] |
| Store payout cycle before and after the advances | 45 days, then weekly | B | Adapty, April 2026. The co-founder: "the money was sitting in Apple's payout queue, and we couldn't touch it for another 45 days." Funds arrived within 48 hours of the application and the company switched to weekly payouts. | [2] |
| Return on the best ad campaigns | over 100% in the first week | B | Adapty, April 2026: "ROI on their best campaigns exceeded 100% in the first week". A vendor's statement about advertising returns, not a payment figure, so graded B. | [2] |
| Attribution cost per install | US$0.04 with Adjust, US$0.02 with Adapty | C | The co-founder's words in the December 2025 case study. Attribution is the matching of an app install to the ad that caused it. | [1] |
| Team | 2 co-founders; 6 people according to a directory | D | Adapty names only "Sergey, co-founder Welmi.ai". RocketReach, a contact-data directory, lists Sergei Tokarev as Co-Founder and CEO, a second co-founder, a head of engineering, a product manager and a creative producer, and writes "Welmi.ai employs 6 employees". The help center articles are signed by the person the directory lists as head of engineering. Not verified by any other source. | [5, 7, 1] |
| Ratings and downloads, October 2026 | 20,000 US App Store ratings at 4.8; over 1 million Google Play downloads and 178,000 reviews | B | Apple's and Google's store pages, read 23 October 2026. The App Store description claims more than 250,000 ratings across countries and the home page claims 226,000 or more; those are the company's own counts. | [3, 4, 8] |
| Android launch | August 2025 | C | AppBrain, a directory of Google Play developers, records T-SHAPED APPS LTD as active since August 2025 with 3 apps and 2 million downloads in total. | [11] |
| Company | T-SHAPED APPS LTD, Griva Digeni 51, Athineon Court, Office 202, 8047 Paphos, Cyprus | B | Seller on both store pages and in the company's own terms. Its App Store developer page lists two more apps, Vea, a period tracker, and Gymtone, a gym workout planner. The UK Companies House register has no company of this name; the Cyprus register's public search needs a login, so the directors could not be read. | [3, 4, 6, 10] |
| Age rating and medical position | 9+ on the App Store; "not a medical device" | B | The store page: "Welmi does not recommend or calculate doses. Follow your doctor's instructions." The help center says the profile's age options start at 18 and the terms require users to be at least 18. | [3, 7, 6] |
| Time from launch to the headline figure | 8 months | calc | Calculated: March 2025 to November 2025. | [1] |
How to read this table: a grade A value is stated by Adapty, the company whose software records the app's subscriptions and runs its ad attribution, in a case study under its own name. A grade B value is a count or a price published by Adapty, Apple or Google that is not a payment figure. A grade C value is the founder's own words or the company's own pages. A grade D value comes from one directory and is not verified. A "calc" value was computed on this page from the rows above it.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free | 0 | 0 | food log; AI analysis results need a subscription |
| Yearly, 4 price points | 2.00 to 5.00 | US$23.99, 35.99, 47.99 or 59.99 a year | |
| Monthly, 3 price points | 11.99 to 19.99 | US$11.99, 14.99 or 19.99 a month | |
| Weekly | 4.99 a week | ||
| premium | US$79.99; period not shown by Apple | ||
| Welmi AI Bundle | US$24.99; period not shown by Apple |
The app is free to download and logs food for free; the store page says that AI analysis results require a subscription. Apple lists four yearly price points, three monthly ones, one weekly one, a product named premium at US$79.99 and a Welmi AI Bundle at US$24.99, whose periods the listing does not show. The monthly equivalents of the yearly prices were computed on this page. The company's own terms say that purchases made on the website go through Stripe and that App Store and Google Play purchases follow the stores' rules.
The timeline
From a launch in March 2025 to more than US$200,000 a month in November 2025, then a reported US$300,000 in 2026, with paid ads as the only named channel. Each entry carries its grade.
- March 2025AWelmi launches on the App Store. Adapty's timeline: "March 2025: Launch". The company starts with Adapty for subscriptions and with Adjust for attribution. [1]
- 3 April 2025CThe earliest archived copy of welmi.ai is a one-page site made on Tilda, a website builder, titled "Calorie Counter & Diet - FitMe" with the line "Personal nutritionist for everyone". [9]
- August 2025AUS$60,000 in monthly recurring revenue. The Android version appears on Google Play the same month, according to AppBrain. [1, 11]
- Second half of 2025CThe company moves attribution from Adjust to Adapty's new attribution product, among the first apps to do so. The co-founder says the per-install price halved, from US$0.04 to US$0.02, which made cheap traffic channels worth testing. [1]
- October 2025AUS$150,000 in monthly recurring revenue. [1]
- November 2025AOver US$200,000 in monthly recurring revenue, eight months after the first organic sale. [1]
- 22 December 2025AAdapty publishes the first case study, "How Welmi.ai scaled from $60K to $200K MRR in 3 months". [1]
- Early 2026BThe company applies for Adapty Finance, which advances up to 85% of unpaid store earnings. Funds arrive within 48 hours and payouts become weekly. The co-founder: "The first week we got a weekly payout through Adapty Finance, we put it straight back into ads." [2]
- 7 April 2026AAdapty publishes the second case study: revenue "from around $30–50K to $300K in about four months", no outside capital, paid acquisition in several regions. [2]
- 5 October 2026BVersion 7.6 on the App Store, after at least 25 versions since April 2026. 20,000 US ratings at 4.8; Google Play shows more than 1 million downloads. [3, 4]
Why it worked
Four causes, in the order they mattered. Each one is a reusable lesson.
Paid ads as the business, not a tactic
No founder audience, no launch site and no creator channel appears in any source. Adapty writes that the team ran "paid acquisition across multiple regions and scaling what worked" and that its best campaigns returned more than 100% in the first week. The product is the paywall and the ad creative as much as the food scanner.
Cheaper attribution opened cheap traffic
Attribution software matches each install to the ad that caused it. With Adjust the company paid about US$0.04 per install, which the co-founder says made low-cost traffic segments unprofitable to test. Adapty's attribution at about US$0.02 per install, connected to the same dashboard as the subscription revenue, let the team see return on ad spend per campaign in real time and test segments it had skipped.
Weekly payouts instead of 45 days
Apple pays developers on a monthly schedule, so money earned in a week was unusable for about 45 days. In 2026 the company took advances on unpaid store earnings from Adapty Finance, for a fixed fee and no equity, and reinvested each week's earnings into the next week's ads. Adapty's second case study attributes the rise to about US$300,000 in about four months to this faster loop.
Photo logging plus a GLP-1 diary
The feature list on the store page is the category standard in 2025: photo, voice and barcode logging, daily calorie and protein targets, meal plans, fasting timer, Apple Health sync. The GLP-1 diary is the addition that follows the market: people on weekly weight-loss injections need to log doses, reminders and side effects, and the app keeps that next to the meals while saying it does not calculate doses.
How it was built
These details come from the store pages, the company's own legal text and website code, and the Adapty case studies. Grades B and C. [3, 4, 1, 2, 6, 8, 9, 7]
- An iPhone app that needs iOS 17.1 or later, 169.8 MB, in English and 16 more languages; an Android version on Google Play since August 2025 under the package name welmi.ai.android.
- Adapty's SDK handles the in-app subscriptions, paywalls and A/B tests; Adapty Attribution replaced Adjust for matching installs to ads; Adapty Finance advances the store payouts. The attribution flow described in the case study takes spend data from Meta Ads and TikTok for Business.
- Purchases on the website go through Stripe, according to the company's privacy policy; the policy also names Google Analytics, Google Tag Manager, Firebase and an AppsFlyer ID, and the app syncs with Apple Health.
- The website is a single-page React app with i18next translations in 17 languages, served from Netlify according to a comment in its HTML; the first site in April 2025 was built on Tilda. The AI model that recognizes food is not named anywhere.
- Support runs on an Intercom help center with 27 articles in 12 languages; the store pages give a support address and a Cyprus phone number.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
The dead that came before
Five companies in the Loot Drop database sold personal nutrition or health insight to consumers and failed. The values below are from that database and are not independently verified.
Refolo
The promiseA meal-planning app with personalized recipes for people moving to a plant-based diet.
Why it diedCharged for meal plans that Pinterest gave away, then larger wellness apps added the same features. Too few users were motivated enough to pay.
Wink Health
The promisePersonal health recommendations on demand, by subscription, from the user's data.
Why it diedCould not hold subscribers against larger health companies and rivals with better data analysis, and regulation added cost.
SpringWit
The promisePersonalized wellness plans and alerts from wearable data, by subscription.
Why it diedApple and Fitbit owned the devices and the data. Generic health tips bored the users, who left.
Airo Health
The promiseA wristband that tracked stress, nutrition intake and exercise and gave real-time advice.
Why it diedCould not justify its price against Fitbit and Apple Watch, had logistics problems with the hardware and ran out of money.
Arivale
The promiseScientific wellness: genome, blood and microbiome tests plus a personal coach, for US$3,500 up front.
Why it diedSpent about US$4,700 per customer to earn about US$1,200 a year from each. The customers lost interest once the novelty of the data passed.
What changed
Each of the five sold the same promise, that personal data about food and the body would lead to better habits, and each needed either hardware, laboratory tests or human coaches, so the cost per customer was high and the device makers owned the data. Refolo, the one pure app, found that people would not pay for meal plans. Welmi's costs are an AI model call per photo and an ad per install; there is no hardware, no test kit and no coach.
Two things in the market changed. First, in 2023 food-photo recognition became good enough that logging a meal takes one photo instead of a search through a database, which removed the chore that made earlier calorie counters lose users. Second, GLP-1 injections turned weight loss into a weekly routine with a dose to record, and an app that logs the dose next to the meals has a reason to be opened every day. Adapty's numbers show that the business grew while the company bought every user with ads, so the question for the next years is whether the subscriptions last longer than the payback period of the ads.
Can one person copy it now?
The product, with difficulty. The money loop, yes.
Food-photo calorie counters are a crowded category: Google Play lists SnapCalorie next to Welmi and Cal AI has its own page on this site. The method that the sources support is the operating loop: measure return on ad spend per campaign with attribution tied to subscription revenue, pay as little as possible per attributed install, and shorten the time between earning and reinvesting. The capital this needs is small but not zero: the company started on 45-day store payouts and later paid a fee for weekly advances.
Where it is still open: Any phone subscription where a weekly routine produces something to log: sleep, medication, training, skin, cycles. The seller already applies the same model to a period tracker and a workout planner.
- 1Pick a subscription app category where a new AI input, such as a photo, removes the main chore of the old apps, and ship the standard feature list around it.
- 2Run paid acquisition from the first month, with attribution connected to subscription revenue so that each campaign's return is visible within days.
- 3Shop for attribution and subscription tooling by price per install; a halved price makes cheap traffic segments testable.
- 4When the payback period is shorter than the store's payout cycle, use payout advances or a credit line to reinvest weekly, and stop when a campaign's first-week return falls below 100%.
What to watch
The risks are the mirror image of the causes above.
The numbers come from a vendor's marketing pages
Adapty publishes case studies to sell its software, and the two Welmi pages do not line up: the December 2025 page gives US$60,000 in August 2025 and over US$200,000 in November 2025, while the April 2026 page quotes a rise from around US$30,000 to 50,000 to US$300,000 in about four months without naming the months. No payment-processor profile, press report or company filing confirms either series. The founder's full name and the headcount of 6 come from one contact-data directory.
Growth bought with ads and payout advances
Every named channel is paid. Growth stops when a campaign's return drops below the advance fee plus the ad cost, and the subscriptions have to outlive the payback period. The store pages show one-week trials and weekly plans, which are the plan types with the shortest lives.
A health app with a GLP-1 diary
The app logs injections of prescription medicines. The store page says that it does not recommend or calculate doses and is not a medical device, and the terms require users to be at least 18 while the App Store rates it 9+. Any change in how Apple, Google or regulators treat apps that touch weight-loss drugs would reach this feature first.
A crowded category with a large competitor
Cal AI and SnapCalorie sell the same photo logging. The AI model behind the recognition is not named and is probably rented from a model provider, so every competitor can buy the same accuracy.
Support complaints in the reviews
The most recent US reviews on the App Store include a paying user locked out of the account with no password recovery and a user whose free trial did not start. The help center is run by one named author.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- A Adapty case study, published 22 December 2025, updated 21 July 2026: How Welmi.ai scaled from $60K to $200K MRR in 3 months with Adapty AttributionRead in full. Adapty's software records the app's subscriptions and runs its attribution. Source for the launch month, the MRR timeline, the Adjust and Adapty per-install prices and the co-founder's first name.
- A Adapty case study, published 7 April 2026: How Welmi grew revenue 6x in 4 months without raising a single dollarRead in full. Source for the US$30,000 to 50,000 to US$300,000 statement, the no-outside-capital statement, the 45-day payout cycle, the first-week returns and Adapty Finance.
- B Apple App Store, US: Welmi - Calorie Counter & Diet, by T-SHAPED APPS LTD, read 23 October 2026Read in full. Source for the feature list, the GLP-1 wording, the age rating, the in-app purchase prices, the ratings count, the version history and the seller.
- B Google Play: Welmi - Calorie Counter & Diet (welmi.ai.android), read 23 October 2026Read the description, the ratings and the developer section. Source for the downloads, the review count, the developer address and phone number, and the SnapCalorie listing among similar apps.
- D RocketReach directory: Welmi.ai management teamRead in full. A contact-data directory. Source for the full name of the co-founder and CEO, the roles of four other people and the count of 6 employees. Not verified elsewhere.
- C Welmi Terms of Use, Privacy Policy and Cancellation & Refund Policy, the English text that welmi.ai/tos, /policy and /cancellation load, read 23 October 2026Read the sections on the provider, age, medical disclaimers, payments, analytics and Apple Health, from the site's English legal text file. Source for the company name and address, the 18-year rule, the Stripe, Google, Firebase and AppsFlyer mentions and the AI Health Reports disclaimer.
- C Welmi Help Center on Intercom, including the article How to set up and adjust your goalsRead the index and one article. Source for the 27 articles in 12 languages, the single named author and the statement that the age options start at 18.
- C Welmi home page, read 23 October 2026Read the HTML and the site's JavaScript bundle, since the page renders by script. Source for the 226,000-plus ratings claim, the company footer and the Netlify, React and i18next details.
- C Wayback Machine copy of welmi.ai from 3 April 2025Read in full. The earliest archived copy: a Tilda page titled Calorie Counter & Diet - FitMe.
- B Apple App Store developer page: T-SHAPED APPS LTDRead in full. Lists three apps: Vea, Gymtone and Welmi.
- C AppBrain developer page: T-SHAPED APPS LTDRead the summary. A directory of Google Play developers. Source for the Android start month, the three apps and the download totals.
- C Loot Drop database export, 1,749 failed startupsLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.