Six calibrated profiles for two devices. Each one is a VDC, an impulse response
and a Liveprog limiter, tuned together and checked before it ships.
Every profile is three files, loaded in RootlessJamesDSP in this order.
| Stage | Job |
|---|---|
1. VDC .vdc |
Biquad banks at 44.1 and 48 kHz, fitted to the same predicted curve |
2. Convolver .irs |
Convolution: 4-channel true-stereo for IEM, 2-channel for speaker |
3. Liveprog .eel |
Input guard, 1.5 ms lookahead limiter, true-peak guard, per-profile controls |
Each chain is normalised to a maximum gain of −0.5 dB, and the limiter protects the peaks. Details are in Method.
KZ EDX Pro · in-ear
| Profile | Character |
|---|---|
| HD | True-stereo four-channel IRS with light crossfeed (−12.5 dB at 500 Hz). 4 controls. |
| Spatial V4 | HRIR-style crossfeed, 0.3 ms inter-aural delay, Binaural Width 50 to 150 %. 5 controls. |
| Reference V10 | Dual-mono, no crossfeed; the VDC carries the correction. 4 controls. |
Redmi 12 · speaker
| Profile | Character |
|---|---|
| Protection V6 | The IRS carries the correction; one 110 Hz high-pass protects the micro-speaker. 5 controls. |
| Cinema V8 | Centre Dialogue Boost and Surround Stage Width on a VDC-led correction. 7 controls. |
| Dynamic V10 | Peak-hold compressor, virtual bass harmonics and loudness makeup. 9 controls. |
Not sure which one? Use the finder or compare all six on the site.
- Download a profile zip below, or the bundle of all six.
- ViPER-DDC: load the profile's
.vdc. - Convolver: load its
.irswith optimisation set to Original. - Liveprog: load its
.eeland leave True Peak Detect on.
Keep the Convolver preamp and the EEL post gain at 0 dB.
Claims come in three tiers that are never mixed.
- Offline, automated. 19 checks on each of 6 builds in
acceptance.json: 108 pass, 0 warn, 6 not run (cpu). - Device-reported. Author-reported listening passes for all six builds are recorded in
device-tests.json. They are not instrument measurements. - Acoustic measurement. None. Every curve is calculated, not measured.
The limits worth knowing: speaker boost is capped at +8 dB, true-peak overshoot is at most +0.25 dB (calculated, 8x oversampling), and no bank exists for sample rates other than 44.1 and 48 kHz. All are listed in Limits.
| Package | Size |
|---|---|
| All six profiles | 73.4 KB |
| KZ EDX Pro · HD | 28.4 KB |
| KZ EDX Pro · Spatial V4 | 3.7 KB |
| KZ EDX Pro · Reference V10 | 4.1 KB |
| Redmi 12 · Protection V6 | 17.1 KB |
| Redmi 12 · Cinema V8 | 10.5 KB |
| Redmi 12 · Dynamic V10 | 11.1 KB |
Changes from the previous release are listed in releases/7.0.0/DIFF.md. To verify the files in releases/7.0.0/:
cd releases && sha256sum -c 7.0.0/SHA256SUMS.txt| Read | For |
|---|---|
| Method | How a profile is built and what the curves mean |
| Evidence | The three evidence tiers and the v6 results |
| Limits | What the profiles do not do, and what would change it |
| Open items | Everything still unfinished, and what each item needs |
| 44.1 kHz refit | The matched-bank refit report |
| Reproduce | Rebuild and verify from source |
| Releasing | Cut and freeze a release |
| Changelog | What changed in each version |
The same docs are rendered on the live site.
HyperAudio/
├── site/ Live web interface (HTML, CSS, JS) and downloads
├── releases/ Frozen, versioned releases with checksums (7.0.0, 6.1.0, 6.0.0, 5.0.0)
├── src/ Source profiles and metadata
├── docs/ Source Markdown, rendered to site/docs/
├── build/ build.py, response.py, docs_site.py, stamp.py, seo.py
├── tests/ check.py, eel/, a11y.js, overflow.js, links.js
├── tools/ Formatter and JS verifier
├── .github/ CI, Pages and release workflows
├── VERSION Current version string
└── index.html Root redirect to /site/
python3 tools/ship.py # regenerate derived files, then gate
python3 tools/ship.py --release # full build, freeze, gate
node tests/overflow.js # layout gates (needs Playwright)
