How we test claims

Audio software is full of big words. We'd rather give you numbers you can check. Every figure on this site comes from one evidence sheet, and every line on it names the measurement, the device, the settings and the source.

Our rules

  1. Measured, or not said. If we haven't measured it, we don't claim it. We haven't measured heat, battery drain or older iPads yet, so you won't find claims about them.
  2. Specific and dated. "5.7 % of one core on an iPad Air M3 13", iPadOS 27.2, 48 kHz, 256-frame buffers, 7 November 2026" instead of "low CPU".
  3. Say what the number is. Our latency figures are the plug-in's own latency, reported to the host, never "round-trip". Our CPU figures say whether they are the audio thread or the whole process.
  4. Fair references. When we compare with another product, we measure both from the outside on the same inputs, name the other product only for the comparison, and say where it does better.
  5. Fix the sheet first. When the product changes, we re-run the measurement and update the evidence before the website.

An example: we corrected ourselves

On 8 November we found that one module reported one sample more latency than it actually delayed. Every latency figure was one sample high. We fixed it, added a host test that checks the output with all modules off is exactly the input delayed by the reported latency, and updated every page. Live mode is 191 samples (4.0 ms), not 192.

How we measure

  • CPU: a benchmark built into the app renders 30 s of a synthetic voice through the plug-in on the iPad and divides the CPU time by the audio time. Raw report · method.
  • Pitch correction: a known-pitch test signal, decision probes and two real takes, scored against a written contract. Method.
  • Noise reduction: real vocals mixed with five noises at two levels, scored on noise removed, voice damage and output SNR. Method.
  • Compressors: static curves and timing measured against reference curves with sine tones and level steps. Method.

What you can do

If a number on this site doesn't match what you measure, tell us on the support page with your device, host and settings. We'll check it, and if we're wrong, we'll fix the page and say so.