Record ~50 seconds of your engine and EngineEar AI tells you what's likely going on β mechanical causes, urgency, and what to check before the workshop. Wear evaluation: does your engine sound its age? The AI compares its acoustic fingerprint to what a car of your make, year and mileage should sound like. Your garage stays free: multiple vehicles, fuel log, maintenance history, cost stats, backup and restore. Only AI analyses are subscription-based. 21 languages.
No reviews yetBe the first to leave a review for EngineEar AI
Maker
π
Hey Product Hunt,
We built EngineEar AI because we were convinced that engine sounds carry more information than we give them credit for. A motor doesn't just "run" β it exhales its condition through vibration, exhaust rhythm, valve ticks, injector timing. Every one of those is a signal, and modern audio models have quietly become good enough to read them.
The feature we're proudest of is wear evaluation: when the AI listens to your engine, it doesn't just look for individual defects β it compares the overall acoustic signature against what a car of your make, model, year, and mileage should sound like. If a 5-year-old car with 50,000 km already sounds like a 200,000 km workhorse, that's worth knowing. Not as a final diagnosis, but as a data point you can bring to your mechanic so you walk in informed instead of guessing.
Everything else is built around that idea. Guided Mode walks you through 30 s of idle + 50 s of revs so the model gets a clean signal. Free Mode lets you record on your own terms. And the whole vehicle side stays free β multiple profiles, fuel log, maintenance schedule with reminders, cost stats, backup and restore β so the AI actually has your vehicle context to compare against. Only the analyses themselves are on subscription.
One honest thing upfront: recognition quality depends on the microphone. Built-in phone mics roll off the low and high frequencies that carry the most diagnostic signal, and background noise amplifies that gap. Because of that variance across phones and environments, the AI errs on the side of flagging β sometimes calling out findings a mechanic would shrug at. In a quiet spot with a decent external mic clipped near the engine, the assessments get sharp. Straight from your phone in a windy parking lot, treat individual line items with a grain of salt and trust the overall trend.
What we'd love your feedback on:
β Try Guided Mode with your own car and tell us if the diagnosis + wear evaluation feel plausible for what you know is going on under the hood.
β Does the "one-time service events" flag in the garage match how you actually track things like warranty extensions or seasonal tire swaps?
β Anything about the result presentation that reduces your trust β missing confidence indicator, jargon, urgency scale?
Thanks for looking, and for the honest takes.
Report
does the ai just guess at what might be wrong or does it actually pull from a database of real engine sound samples for my specific model? curious how confident the diagnoses actually are before i pay for a subscription.
thank you for your Question. What the AI does here is pattern recognition. Depending on the microphone of the smartphone and environmental noises this can be good or less accurate. We actually configured it that way, that it'll tend to notify more, where might be less, just to cover the possible environmental circumstances , and the input quality loss through a smartphone microphone.
Report
How accurate is the AI actually for weird noises like timing chain rattle versus a worn tensioner? Would love to see what kind of confidence score it gives before I trust it over my mechanic.
Report
Maker
@ezel656461Β Honest answer on timing chain rattle vs. worn tensioner: these two are one of the harder acoustic pairs to separate cleanly. Both produce metallic rattle in overlapping frequency bands, and the tell β tensioner is often cold-start / low-load, chain wear tends to broaden across RPM β doesn't always show up in a single 50-second recording. The AI usually returns both as ranked mechanical causes rather than committing to one, which I think is the honest output for that specific pair.
On the confidence score β yes, every result includes a Confidence badge (High / Medium / Low) that reflects how cleanly the acoustic signature mapped to known patterns. Low doesn't mean "wrong"; it means "the audio is ambiguous or noisy enough that you should treat the ranking as a hypothesis list, not a verdict." That's why the app is framed as something to walk into the workshop with, not something to replace the workshop.
If you try it, one tip: record in Guided Mode (30 s idle + 20 s revs), and if you can, do one recording cold-start and one warm. The delta between them is genuinely useful for narrowing down chain vs. tensioner β and a good data point to bring to your mechanic either way. Would love to hear whether the ranking matches what they end up finding.
does the ai just guess at what might be wrong or does it actually pull from a database of real engine sound samples for my specific model? curious how confident the diagnoses actually are before i pay for a subscription.
@adem132916Β Hi Adem,
thank you for your Question. What the AI does here is pattern recognition. Depending on the microphone of the smartphone and environmental noises this can be good or less accurate. We actually configured it that way, that it'll tend to notify more, where might be less, just to cover the possible environmental circumstances , and the input quality loss through a smartphone microphone.
How accurate is the AI actually for weird noises like timing chain rattle versus a worn tensioner? Would love to see what kind of confidence score it gives before I trust it over my mechanic.
@ezel656461Β Honest answer on timing chain rattle vs. worn tensioner: these two are one of the harder acoustic pairs to separate cleanly. Both produce metallic rattle in overlapping frequency bands, and the tell β tensioner is often cold-start / low-load, chain wear tends to broaden across RPM β doesn't always show up in a single 50-second recording. The AI usually returns both as ranked mechanical causes rather than committing to one, which I think is the honest output for that specific pair.
On the confidence score β yes, every result includes a Confidence badge (High / Medium / Low) that reflects how cleanly the acoustic signature mapped to known patterns. Low doesn't mean "wrong"; it means "the audio is ambiguous or noisy enough that you should treat the ranking as a hypothesis list, not a verdict." That's why the app is framed as something to walk into the workshop with, not something to replace the workshop.
If you try it, one tip: record in Guided Mode (30 s idle + 20 s revs), and if you can, do one recording cold-start and one warm. The delta between them is genuinely useful for narrowing down chain vs. tensioner β and a good data point to bring to your mechanic either way. Would love to hear whether the ranking matches what they end up finding.