On-device vs cloud audio cleaning
Why on-device voice denoise matters: no upload, no account, no app telemetry. How ARP’s local RNNoise path differs from cloud AI enhancers.
What cloud cleaning usually means
Most “AI audio cleaning” products upload your file to a server, run a model, and send back a result. That can sound excellent. It also means a third party receives the recording — even briefly.
What on-device means in ARP
ARP cleans (RNNoise plus light shaping) and transcribes (Apple's on-device speech recognition) on the iPhone or iPad that holds the file. Audio is not uploaded — ARP has no servers. No ARP account. The app has no analytics SDKs. The one network request is Apple's speech model, downloaded once the first time you transcribe.
The marketing site may use PostHog for website analytics. That script is not in the apps. See Privacy Policy.
Honest tradeoffs
- Privacy and control — the file never leaves for ARP’s processing.
- Capability — cloud services can ship larger models and batch pipelines. ARP processes one track at a time.
- Offline — once installed, cleaning does not need a network path for the audio itself.
When to choose which
Choose cloud if you need maximum model horsepower and are comfortable uploading. Choose ARP when the job is “strip the room from this memo/interview/lecture on this device.” Related: voice memo, interview, lecture.
Your recording stays on your device. 10 recordings free; Premium is a one-time purchase.
