Repository Public License
1<!--2ORIX Audio3File: docs/reports/performance.md4Author: Simon-Pierre Boucher5Contact: contact@spboucher.ai6-->78# Performance Report910**Date:** 2026-11-20 · **Tool:** `orix-bench` (release build of the core package) · **Machine:** Mac, Apple M5 Max, macOS 27 · **Threads:** one1112> **These are not iPhone measurements.** They show what the algorithms cost on a fast desktop-class chip. An iPhone will be slower by a factor that has not been measured, and nothing here speaks to battery, heat or memory pressure on a phone.1314Raw output: `Benchmarks/results/dsp-throughput.md`.1516## Signal processing1718| Operation | Result |19|---|---|20| FFT 1024 — window, transform, scaling | 0.8 µs |21| FFT 4096 | 2.5 µs |22| FFT 16 384 | 12.8 µs |23| Live spectrum frame — 4096-point FFT, smoothing, peak hold, seven bands | 15.1 µs |24| Waveform summary with band colours, 60 s of 48 kHz audio | 14 ms (4433× real time) |25| Spectrogram, FFT 2048, hop 512, 60 s (5622 columns) | 15 ms (3904× real time) |26| Rasterise a 512 × 512 spectrogram tile through a palette table | 0.65 ms |27| BS.1770 loudness with true peak, 60 s mono | 95 ms (634× real time) |2829Timings vary by a few percent between runs; an earlier run of the same binary gave 0.61 ms for the tile and 12 ms for the waveform summary.3031## Whole files3233Ten minutes of 48 kHz stereo pink noise, 24-bit source.3435| Operation | Time | Speed | Output size |36|---|---|---|---|37| Full analysis pass (levels, loudness, spectrum, speech activity, waveform) | 2.74 s | 219× | — |38| Export to WAV, PCM 16-bit | 0.10 s | 5739× | 115.2 MB |39| Export to Apple Lossless | 1.28 s | 470× | 104.6 MB |40| Export to AAC | 2.11 s | 285× | 8.7 MB |41| Noise reduction + voice enhance + export to WAV | 3.67 s | 163× | — |4243(Pink noise is close to incompressible, which is why Apple Lossless saves little here; speech and music compress much better.)4445## Transcription and notes4647| Operation | Result |48|---|---|49| Transcribe about 60 s of speech (`SpeechTranscriber`, model installed) | 0.59–0.70 s, 82–116× real time |50| Generate notes from a one-minute transcript (Foundation Models) | about 7–8 s |51| First-time speech model download | several minutes, once per language |5253## What these numbers support5455- **Canvas and Core Graphics instead of Metal.** A spectrum frame costs microseconds and a spectrogram tile is rasterised once in well under a millisecond, against a 16.7 ms frame. See [visualization](../architecture/visualization.md).56- **Analysing every recording automatically.** A ten-minute file takes under three seconds here.57- **Encoding on stop rather than while recording.** Encoding is hundreds of times faster than real time, so finishing a long session is a short wait.5859## Simulator observations6061Not benchmarks, but worth recording:6263- In the iOS 26.5 simulator (Debug build) the analysis thread produced every expected waveform bucket and spectrogram column while capturing 48 kHz stereo and 22.05 kHz mono; this is asserted in the app tests.64- No samples were dropped in any test run.6566## Not measured6768| Item | Why |69|---|---|70| CPU while recording on an iPhone | no device |71| Battery drain per hour, screen on and locked | no device |72| Memory over a multi-hour recording | no device; by design only the waveform summary grows with duration (about 13.5 MB per hour at 48 kHz) |73| Frame rate of the live views on older iPhones | no device |74| Long-file transcription throughput | only one-minute samples were run |75| Thermal behaviour | no device |7677No figure is given for any of these, and none should be inferred from the Mac results.78