Optiforge-Max

Android system-wide 5-band audio compressor/limiter (EQ + dynamics) — eu.youforgemax.optiforgemax.

5-band compressor / limiter for Android tablets. Two modes, switchable from the top tab bar:

• Live — system-wide realtime processing of all output audio.
• Audio File — offline open → preview → save for an audio (or video) file.

No root, no NDK. Requires Android 10+ (minSdk 29). Signed release APKs are on GitHub Releases.

📖 End-user guide: MANUAL.md — step-by-step for both modes, control reference, calibration, troubleshooting. This README is the technical overview. Version history: CHANGELOG.md.

---

Live mode

Built on android.media.audiofx.DynamicsProcessing (API 28+). No audio capture. The effect attaches to the global output mix session (session 0), so it processes all system audio.

Controls (per spec)

• Input gain — pre-everything, −40…+40 dB
• Per band (×5): threshold (−100…0 dBFS), attack (0…500 ms), release (0…3000 ms), ratio (1…20:1), makeup (−12…+24 dB), enable
• Output limiter: threshold/attack/release/ratio/post-gain
• Crossover cutoffs default 150 / 600 / 2500 / 7000 / 20000 Hz (editable in code)
• Per-band gain-reduction meters (needs RECORD_AUDIO)
• Spectrum view — 32 log-spaced bars from the same FFT (bar-centric folding, so low bands aren't dropped when a bar is narrower than one FFT bin)
• Link attack/release — when on, moving any band's Attack or Release writes all 5 bands at once; Attack/Release stay independent, limiter untouched
• Presets — named save/load/delete (SharedPreferences, JSON)
• Meter calibration slider — tune the estimated level offset on-device

Gain-reduction meters — how / caveat

DynamicsProcessing exposes no GR readout. Meters are estimated: a Visualizer on the same session FFTs the output mix, bins are folded into the 5 bands to estimate each band's level (dBFS), then run through the soft-knee compression curve (DspEngine.estimateGrDb) with attack/release smoothing. Tracks compression activity for visual feedback — not a calibrated measurement (post-mix magnitude, not the compressor's internal detector; makeup/limiter not reflected). Denied mic permission disables meters; everything else still works.

Calibration: the byte-magnitude→dBFS mapping (ref 128) is approximate and varies per device. With audio playing below threshold, nudge the Meter calibration slider until GR sits near 0, then set thresholds to taste. The offset is saved in presets.

All ranges are UI-only — widen them in MainActivity.kt (SliderRow calls). The underlying floats accept far wider values.

---

Audio File mode

Process a file offline and save the result — same 5-band sound as Live mode, rendered in pure Kotlin on PCM (the live DynamicsProcessing effect has no file-render API).

1. Open — pick an audio or video file. For a video, the audio track is extracted and processed; the video is discarded.
2. Preview — loops a 60 s window (start point selectable) through the exact render chain. Every slider edit is audible live.
3. Save — renders the whole file to an AAC .m4a via the system codec. Runs in the background with progress + cancel.

Signal chain (audiofile/ package):

input gain -> LR4 5-band crossover -> per-band soft-knee compressor + makeup
           -> sum -> limiter (+ brickwall ceiling) -> [optional loudness gain]

• Loudness normalize — optional pass measures integrated loudness (simplified ITU-R BS.1770 K-weighting) and applies a single gain to hit a LUFS target (default −14), clamped to ±24 dB.
• Has its own preset store (presets_audiofile) plus per-band knee and the same Link attack/release toggle.
• Whole encoded stream is buffered in memory before muxing — fine for typical clips, heavy for multi-hour files.

Build

SDK on the dev box at ~/android-sdk. The Gradle wrapper jar is not committed; either open in Android Studio (Koala+) to generate it, or build with a local Gradle 8.9:

# debug
gradle :app:assembleDebug

# signed release (keystore props or env: RELEASE_STORE_FILE / _PASSWORD /
# RELEASE_KEY_ALIAS / RELEASE_KEY_PASSWORD; falls back to unsigned if absent)
gradle :app:assembleRelease \
  -PRELEASE_STORE_FILE=/path/to/optiforgemax.jks \
  -PRELEASE_STORE_PASSWORD=*** -PRELEASE_KEY_ALIAS=optiforgemax \
  -PRELEASE_KEY_PASSWORD=***

Caveat — global session & per-player binding

Stock Android honors output-mix effects, but many players bypass the global session-0 effect: VLC (its own audio session + non-mixer output), and some OEM paths (Samsung/Xiaomi, offloaded/MQA/Dolby) for compressed/offloaded playback.

Since 1.5 the app registers a receiver for the standard OPEN/CLOSE_AUDIO_EFFECT_CONTROL_SESSION broadcasts and automatically binds the compressor to a player's announced session id (DspEngine.attach(id) + re-applies state + rebinds the meter). The Routing card shows the current binding; "Use global mix" reverts to session 0.

For a player to be bound it must announce its session:

• VLC: enable Audio → Audio effects (and/or set output to AudioTrack).
• Offloaded audio: disable "MQA/Dolby/audio offload" in the player or Developer Options so the chain stays in the software mixer.
• If nothing announces a session, the effect stays on the global mix; use Retry attach if attach failed.