Three AI wrappers at $2K–$10.9K MRR: boxing, abs, and launch videos (2026)

[data] Three products qualify for this week's teardown: Boxer AI, AbMaxx, and Motionfly. TrustMRR lists founding months from March to April 2026, so each sits inside the channel's 12-month window. The dates are public founding dates, not exact App Store launch dates. Boxer AI is the camera-based boxing coach with $8,414 MRR and 673 active subscriptions. AbMaxx turns an abs photo into a tailored routine and lists $10,924 MRR, although its snapshot shows no active subscriptions. Motionfly turns one product description into a launch video and lists $2,016 MRR with 50 active subscriptions. 1 2 3

TL;DR

[data] Boxer AI and AbMaxx sell personalization through a phone camera. Boxer tracks punches and gives live technique feedback; AbMaxx analyzes an abs photo and builds a workout plan. Motionfly sells finished media: describe a product and receive a launch video with mockups, animations, voiceover, and sound. Public TrustMRR snapshots put their MRR at $8,414, $10,924, and $2,016, respectively. 1 2 3

[editor's view] Boxer AI has the clearest repeat-use loop because every workout can produce another score. AbMaxx has the strongest revenue snapshot but also the largest evidence wrinkle: the same page lists no active subscriptions. Motionfly is the easiest manual service to test, yet the final product has to keep script, visuals, voice, and pacing coherent. The cloneable layer is the workflow around the model. The hard layer is trust, repeat use, and distribution.

Three takeaways

[data] Boxer AI lists a one-person team, 673 active subscriptions, and TikTok, Instagram, and TikTok Ads as marketing channels. 1
[data] AbMaxx lists $41,414 in last-30-day revenue and $10,924 MRR, while its page shows no active subscriptions and names TikTok, Instagram, and Meta Ads as channels. 2
[data] Motionfly lists 50 active subscriptions, $2,016 MRR, and X/Twitter plus Reddit as channels; the public pages do not attribute revenue to one primary channel or show a paid-plan price. 3 4

Replication scoreboard

[editor's view] More stars mean harder to replicate. For legal risk, more stars mean more exposure rather than better legal hygiene. These are editorial estimates based on the public product surfaces and August 24, 2026 TrustMRR snapshots. Boxer AI's page shows a team size of one. AbMaxx and Motionfly each name one founder, while their pages do not expose a formal team-size field. 1 2 3

Product	Technical lift	Information edge	Capital needed	Legal risk	Clone read
Boxer AI	★★★★☆	★★★☆☆	★★★☆☆	★★★★☆	[editor's view] The camera loop is buildable; feedback quality and user trust are harder
AbMaxx	★★★☆☆	★★★☆☆	★★☆☆☆	★★★★★	[editor's view] The scan-to-plan funnel is simple to demo; health-data handling raises the floor
Motionfly	★★★★☆	★★☆☆☆	★★★☆☆	★★★☆☆	[editor's view] Prompt-to-video is easy to show; coherent output and unit economics decide the business

Boxer AI: the workout is the data loop

[data] Positioning: Boxer AI turns an iPhone into a personal boxing coach. The app uses the camera to track jabs, crosses, hooks, and uppercuts, then gives live coaching and post-session feedback. The job-to-be-done is specific: help a boxer practice technique at home and see whether the next session improved. 5

[data] Traction and team: TrustMRR shows $8,414 MRR, $9,250 in last-30-day revenue, 673 active subscriptions, and RevenueCat verification as of August 24, 2026. The listing names Ben Marlowe as founder, shows a March 2026 founding date, and states a one-person team. 1 The App Store page shows 4.7 stars from roughly 1.5K ratings and lists the app as free with in-app purchases. 5

[data] Acquisition, pricing, and stack: TrustMRR lists Instagram, TikTok, and TikTok Ads. The public listing does not show revenue attribution by channel, so those are reported acquisition surfaces rather than a proven primary source. The App Store lists subscription price points from $8.99 to $34.99. TrustMRR lists SwiftUI and Swift on the front end, plus Python, FastAPI, Google Cloud, OpenAI, Firebase, and RevenueCat on the back end. 1 5

[editor's view] The model call is only one layer of the product. A clone needs camera capture that works in a living room, punch detection that users believe, feedback that changes the next drill, and a progress screen that creates a reason to return. The 673 active subscriptions give Boxer AI a stronger repeat-use signal than a one-time image generator. The public data still says little about retention or how many sessions a subscriber completes each month.

[editor's view] Replication: technical lift ★★★★☆; information edge ★★★☆☆; capital needed ★★★☆☆; legal risk ★★★★☆. The information edge could grow from labeled training sessions and the drills that help users fix recurring mistakes. The legal risk comes from camera footage, user-generated content, age limits, and fitness claims. The App Store says Boxer AI is for fitness and technique practice, and that it does not replace professional instruction or medical advice. 5

If you wanted to copy this

[editor's view] First concrete step: build a phone-camera prototype for one punch combination and test whether ten boxers accept its feedback after three sessions. First likely failure mode: the app produces a confident score that users can see is wrong, so the coaching loop loses credibility before subscription conversion matters.

AbMaxx: the scan is the hook, the plan is the product

[data] Positioning: AbMaxx asks users to take a photo of their abs, analyzes the image, and creates a routine aimed at the user's weak areas. The job-to-be-done is to remove the guesswork from choosing ab exercises and give a user a simple plan they can repeat. The App Store description says its analysis evaluates 1,000 core data points and adapts the routine to the user's physique. 6

[data] Traction and team: TrustMRR shows $10,924 MRR and $41,414 in last-30-day RevenueCat revenue as of August 24, 2026. The same snapshot shows no active subscriptions. The listing names Lino as founder, gives a March 2026 founding date, and carries the founder's claim that the app launched three months earlier with a 12% download-to-paid conversion rate. The public page does not expose a formal team-size field. 2 The App Store lists 4.7 stars from 875 ratings and in-app prices including $11.99 monthly, $34.99 yearly, and weekly options. 6

[editor's view] The revenue snapshot deserves two readings. The dollar figures show that the funnel has generated real paid demand. The zero active-subscription field makes recurring retention hard to judge, especially because the product sells subscriptions. An indie developer should ask for cohort behavior before treating the MRR figure as proof of a durable fitness business.

[data] Acquisition, pricing, and stack: TrustMRR lists Instagram, TikTok, and Meta Ads, but the page does not identify one primary channel or attach revenue to a channel. It lists SwiftUI and RevenueCat. The App Store says the product may collect contact information, usage data, health and fitness data, purchases, and user content, with some data used for tracking. 2 6

[editor's view] AbMaxx is easier to reproduce than Boxer AI at the first-demo level. A photo upload, a vision model, a structured routine, and a paywall can produce a convincing prototype. The difficult question is whether the scan changes the plan in a way users can feel. A generic workout generator with an abs-themed camera screen will have a low repeat-use ceiling.

[editor's view] Replication: technical lift ★★★☆☆; information edge ★★★☆☆; capital needed ★★☆☆☆; legal risk ★★★★★. The information edge could come from the link between a user's scan, completed routines, and visible progress. The legal risk is high because the product handles body images and health-related information while making personalized fitness claims. A clone needs explicit consent, careful retention rules, and language that avoids turning a fitness estimate into a medical conclusion. 6

If you wanted to copy this

[editor's view] First concrete step: manually review 30 consented photos and deliver one narrowly defined ab routine with a written explanation of what the image can and cannot tell you. First likely failure mode: users like the scan as a novelty but stop after the first plan because the routine feels interchangeable with a free workout search.

Motionfly: the wrapper ends where the video begins

[data] Positioning: Motionfly turns a product description into a launch video with 3D device mockups, text animation, voiceover, music, and sound effects. The job-to-be-done is to help a founder publish a credible launch or SaaS explainer without learning After Effects, Premiere, or a motion-design workflow. 4

[data] Traction and team: TrustMRR shows $2,016 MRR, $7,532 in all-time revenue, and 50 active subscriptions, with Stripe verification and a last update on August 24, 2026. The listing names NATE G as founder and gives an April 2026 founding date. The public page does not expose a formal team-size field. 3

[data] Acquisition, pricing, and stack: TrustMRR lists X/Twitter and Reddit as marketing channels, without proving which channel supplies most paid users. The official site offers a free starting point and names use cases such as Product Hunt launch videos, SaaS explainers, app launch videos, and App Store previews. The site does not publish a paid-plan price in the retrieved page. TrustMRR lists React, Node.js, and Stripe. 3 4

[editor's view] Motionfly has a clean manual-service test. A founder can send a customer one intake form, produce one 30-to-60-second video, and learn whether the customer values speed enough to pay. The software becomes harder when the output must match a product's real UI, use a voice the founder approves, fit multiple aspect ratios, and avoid the visual mistakes that make an AI video look generic. The site claims most videos take under 60 seconds to generate, but the public page does not show the cost per render or the share of generations that need manual repair. 4

[editor's view] Replication: technical lift ★★★★☆; information edge ★★☆☆☆; capital needed ★★★☆☆; legal risk ★★★☆☆. The underlying model providers are accessible. The information edge has to come from templates that fit launch contexts, product screenshots that stay accurate, and examples that convert. Capital needs rise with voice, video, storage, and failed-generation costs. Legal exposure includes third-party logos, copyrighted music, cloned voices, and customer claims inside generated marketing material.

If you wanted to copy this

[editor's view] First concrete step: offer five founders a manual launch-video package with one fixed format, one voice, and one revision round. First likely failure mode: the first render looks impressive in isolation but misrepresents the product, forcing so much manual correction that the wrapper cannot support its advertised price.

What these three wrappers are really selling

[editor's view] Boxer AI, AbMaxx, and Motionfly all put a model behind a narrow promise, yet the repeated value comes from the layer after generation. Boxer AI asks whether the next workout changes because of the feedback. AbMaxx asks whether a body scan leads to a plan that a user follows. Motionfly asks whether a founder ships a better launch video faster than a freelancer or a template can deliver.

[editor's view] The safest clone test follows the same order: sell the human-assisted result first, measure repeat use second, and automate the expensive step third. Boxer AI needs trusted feedback. AbMaxx needs a plan that earns a second session. Motionfly needs a video that survives customer review. Model access is the starting input in all three cases; it is not the reason a user keeps paying.

What to do tomorrow morning

[editor's view] Pick one narrow job, recruit five real users, and deliver the result manually before writing the wrapper. Ask each user to pay for a second use. The answer to that second-use question will tell you whether the opportunity is a product, a service, or a demo.

Sources

[data] The links below are the original product, App Store, and revenue pages used for the claims in this issue.

Source	Used for
[data] Boxer AI on TrustMRR	Revenue snapshot, active subscriptions, founder, founding month, team field, channels, and stack
[data] Boxer AI on the App Store	Camera workflow, pricing, ratings, fitness disclaimer, and privacy summary
[data] AbMaxx on TrustMRR	Revenue snapshot, founder, founding month, channel list, stack, and founder conversion claim
[data] AbMaxx on the App Store	Photo-analysis workflow, pricing, ratings, health-data categories, and privacy summary
[data] Motionfly on TrustMRR	Revenue snapshot, active subscriptions, founder, founding month, channels, and stack
[data] Motionfly official site	Prompt-to-video workflow, output types, free entry point, and generation claims

References

1
Boxer AI on TrustMRRtrustmrr.com
2
AbMaxx on TrustMRRtrustmrr.com
3
Motionfly on TrustMRRtrustmrr.com
4
Motionfly official sitemotionfly.co
5
Boxer AI on the App Storeapps.apple.com
6
AbMaxx on the App Storeapps.apple.com

TL;DR

Three takeaways

Replication scoreboard

Boxer AI: the workout is the data loop

If you wanted to copy this

AbMaxx: the scan is the hook, the plan is the product

If you wanted to copy this

Motionfly: the wrapper ends where the video begins

If you wanted to copy this

What these three wrappers are really selling

What to do tomorrow morning

Sources

References

Related content