I'm Akshay Sarode, founder of AceSense. In 2023, I almost bought an iPhone for the sole purpose of using SwingVision. I'm an Android user. I play amateur tennis (NTRP ~3.5 at the time). I had been watching the SwingVision YouTube videos for months — the 25k-view honest review, the unbiased one, the on-court Apple Watch demos. SwingVision looked like the answer to the question every amateur player asks: what am I actually doing wrong? The catch: it was iOS-only, and a Reddit thread titled "Genuinely thinking of getting a iPhone just for the swing[vision]" (r/10s) summed up exactly what was on my mind. Here's what changed it.
Bottom line up front
- I almost spent ~€900 on an iPhone to run SwingVision on.
- I priced out year-one cost: phone + SwingVision tier + Apple Watch and it crossed €1,500 fast.
- I tried the alternative path — building tennis AI on Android — and it worked.
- That alternative became AceSense. This is the story of why.
The Saturday I came close
October 2023. Saturday afternoon. I'd lost a club match 6-3, 6-4 to a guy I'd beaten the month before. Couldn't tell you why. My forehand felt off, my serve felt off, the score said it was off — I had no diagnostic. The match was on a club camera so I had film. I sat on my couch, opened the laptop, scrubbed back through 90 minutes of video, and after maybe ten minutes I gave up. I couldn't see what was happening from a wide-angle court video any better than I could see it from inside my own head during the match. I needed something to tell me where the leak was.
That night I almost ordered an iPhone 15. I had the cart open. The math was: phone €899, SwingVision Pro $24.99/mo (per SourceForge pricing), and the constant chatter from reviewers that the real features lived at the $39.99/mo Max tier (Tennisnerd's review describes the same pricing wall: "the $150/year plan gets you HD recording, but 4K and more accurate line calling require paying $400 annually"). Add an Apple Watch SE for line calling and we're at year-one cost north of €1,500. For one tennis app.
I closed the cart. Then I read the Reddit thread.
The thread that did it
The r/10s post is short, but the comments are not. People are negotiating with themselves out loud:
- "I have an Android. I'm seriously considering switching to iOS just for this."
- "It's the only tennis app worth using. I can't believe it doesn't exist on Android."
- "Does anyone know if there's an Android alternative?"
The Talk Tennis thread "Will SwingVision still work with an Android" (forum link) is the calmer version of the same question, and the answer hasn't moved in years: no. SwingVision's own Android update newsletter is the polite corporate version: we're exploring it. No date.
I'm an engineer. The thing that bothered me wasn't that SwingVision was iOS-only. It was that the entire amateur tennis world was being told "buy an iPhone or wait." That's not a tech problem. That's a business decision a competitor could un-make.
What I tried first
Before I built anything, I tried every Android-compatible option I could find:
- OnForm (pricing, $9.99–$59.99/mo). Worked on Android. Beautiful video annotation tool. But: no tennis AI. I had to scrub through the video and tag shots myself. Same problem as the laptop.
- A generic coach-on-video service (TennisPal-style). Submit video, wait 3–5 days for a human review, pay per video. The latency killed it. By Tuesday's lesson the coach hadn't replied to Saturday's video.
- TennisAI.net (€15/mo). Working Android app. Useful charting. But shallower stroke-quality output than what I'd seen SwingVision do.
- BaselineTennisAI. Working Android app. Vendor diversity option. Fine but not deep enough for the failure-mode question I was trying to answer.
None of them gave me the loop I wanted: record Saturday → tagged report Saturday afternoon → into Tuesday's lesson. So I started prototyping.
The actual build (short version)
My day job background is in computer vision. I started with the same components anyone in this space would: a TrackNet-style ball detector for the small fast-moving target, court keypoint detection because everything is downstream of knowing where the lines are, MediaPipe pose for the player's body, FasterRCNN for player bounding boxes, and a CatBoost model on top of pose features and bounce timing to classify shots and call bounces. The whole pipeline is in /features/shot-detection if you want the specifics; the source-of-truth diagram is in our docs.
The first prototype ran on a single tennis match I'd recorded on a Pixel 6. It misclassified about half the volleys, missed a chunk of second serves, and put the court line keypoints in the wrong place on a clay court. I knew this because I built a regression test (compare_events.py) that compares model output to a hand-annotated ground-truth file. Most apps don't publish their accuracy because most apps don't measure it. I measured it because I needed to know whether I was kidding myself.
Six months later, I had something that produced a per-shot report I'd actually look at on a Tuesday morning. That was the AceSense MVP.
The cost comparison, finally
Here's the spreadsheet I should have done before I almost ordered the iPhone.
| Path | Year-one cost | Devices required |
|---|---|---|
| iPhone + SwingVision Pro | ~€1,160 (€899 phone + $24.99/mo × 12) | iPhone, optional Apple Watch (€280) |
| iPhone + SwingVision Max | ~€1,340 ($39.99/mo × 12) | iPhone, Apple Watch recommended |
| Android (already own) + AceSense | €0–€180 (free tier or Pro) | Phone you already own + €25 fence clip |
The €1,000+ delta isn't a discount on the same product. It's the cost of the iOS lock-in being load-bearing for the SwingVision business model. You're paying for the Apple Watch line-calling experience, the long-baked iPad chrome, the brand. If those three things are what you actually need, that price is fair. For me, they weren't.
Where SwingVision is still the right call
I want to be clear: if you already own an iPhone and an Apple Watch and you want real-time on-court line calling, SwingVision is excellent and AceSense doesn't try to compete on that axis. We don't ship an Apple Watch app. Our analysis is post-match, not in-the-moment. That's a deliberate trade — pose-based stroke quality is hard to do well in real time on a wrist computer, and the per-shot diagnostic loop I wanted didn't need it.
If your sport is more racket-and-paddle adjacent than tennis, SwingVision's coverage there is also still ahead of ours.
For everybody else — Android owners, EU players, people who balked at the $400/yr complaint — that's the gap AceSense is built for.
The lesson from the cart abandonment
I keep that closed iPhone cart screenshot somewhere. It reminds me of two things.
One: the most expensive thing in software is a vendor decision the user can't un-make. SwingVision's iOS-only choice was rational in 2019 — Apple's CoreML and Apple Watch APIs gave them a real product wedge. But by 2023 it was an open wound for the Android half of the amateur tennis world.
Two: the right answer to "should I buy an iPhone for SwingVision" is almost always no. Buy an iPhone if you want an iPhone for general reasons. Don't buy an iPhone for one app, especially not when there's a working Android alternative for the actual loop you want.
What I'd tell my October-2023 self
If you're standing where I was — Android player, considering an iPhone purchase to run SwingVision, doing the YouTube-review death-scroll at 11pm — try this first:
- Try AceSense free on your Android phone. One Saturday match. Zero hardware purchases.
- Read the side-by-side comparison between the two products. Decide which axes actually matter to you.
- Then decide whether the iPhone is worth €900 to get the axes AceSense doesn't cover (Apple Watch line calling, etc.).
Most of the time, you'll skip step 3.
FAQ
Is buying an iPhone just for SwingVision worth it? For most amateur players, no. Year-one cost is roughly €1,200–€1,500. AceSense gives you the same core pipeline on the Android phone you already own.
Why did SwingVision skip Android? They built deeply on iOS-specific APIs (CoreML, Apple Watch). Re-architecting for Android is real engineering work — see their Android update.
Did you build AceSense because of SwingVision? Partly. The Reddit thread "Genuinely thinking of getting a iPhone just for the swing[vision]" was one of several signals that the iOS lock-in was a real problem for real players.
What hardware do I need for AceSense? A phone you already own and a fence clip or basic tripod ($20–$30). No Apple Watch, no proprietary camera.
Akshay Sarode is the founder of AceSense. If this story sounds familiar, start free on Android or read The SwingVision Android alternative, explained.