diff options
Diffstat (limited to 'config.toml')
| -rw-r--r-- | config.toml | 91 |
1 files changed, 91 insertions, 0 deletions
diff --git a/config.toml b/config.toml new file mode 100644 index 0000000..412ba17 --- /dev/null +++ b/config.toml @@ -0,0 +1,91 @@ +# producer settings — CLI flags always override anything set here. + +# "audiobook" = close-narration voice chain, RMS -20 dB, true peak <= -3 dB, 44.1 kHz +# "podcast" = broadcast chain, -16 LUFS, true peak <= -1.5 dBTP, 48 kHz, brighter EQ +# "radio" = deep, warm broadcast voice: +3 dB low shelf @ 100 Hz, tighter mud cut, +# heavy compression, tape saturation; -16 LUFS, 48 kHz +profile = "radio" + +# Output container/codec: wav | flac | mp3 (default: wav) +# format = "wav" + +# Bit depth: 16 | 24 | 32 (default: 16) +# bit_depth = 16 + +# Output sample rate in Hz (default: 44100 audiobook / 48000 podcast) +# sample_rate = 44100 + +# Compute device: auto (CUDA if present) | cuda | cpu +# device = "auto" + +# Loudness target and true-peak ceiling in dB. +# Without these, each profile uses RMS -20 / ceiling -3 (audiobook) or -16 LUFS / -1.5 (podcast). +# target = -20.0 +# ceiling = -3.0 + +# How much of the denoised/enhanced signal to blend in (0-1). 0 disables the stage. +# denoise_strength = 1.0 +# enhance_strength = 1.0 + +# Denoiser: "dfn3" (DeepFilterNet3, default, fast, 48 kHz full-band) +# "zipenhancer" (ICASSP 2025 SOTA, 16 kHz native, bandwidth restored) +# "off" (pure DSP, no heavy downloads) +# Heavy deps + model weights are installed lazily on first use. +[denoise] +engine = "dfn3" +strength = 1.0 + +# Speech enhancement: "off" (default), "mossformer2" (studio restoration), +# "resemble" (generative, opt-in isolated venv, may alter timbre) +[enhance] +engine = "off" + +# Per-profile voice DSP strengths, 0-1 each (CLI flags --hpf, --mud, ... always win). +# 0 disables a stage entirely; every stage is optional. +# hpf = strength of the high-pass (0 turns it off; corner set via hpf_hz) +# mud = cut around 300 Hz (boxy mud) +# warmth = low-shelf lift (position set per profile: 150 Hz audiobook/podcast, 100 Hz radio) +# soothe = dynamic EQ that ducks boxy (200-450 Hz) and harsh (2.5-6 kHz) resonances +# only while they stick out — unlike static EQ, "off" when the voice is clean +# compress = serial 2:1 + 3:1 compression +# tape = asymmetric soft-clip saturation — analog/even-harmonic radio warmth +# deess = de-esser active around 5.5-8 kHz +# presence = boost around 3 kHz +# air = high shelf from 9-10 kHz +# breath = expander depth used to duck breaths + +[audiobook] +mud = 0.8 +warmth = 0.8 +soothe = 0.3 +compress = 0.8 +tape = 0.0 +deess = 0.6 +presence = 0.8 +air = 0.6 +breath = 0.35 +# hpf_hz = 80.0 # high-pass corner frequency (audiobook default 80) + +[podcast] +mud = 0.6 +warmth = 0.7 +soothe = 0.5 +compress = 0.9 +tape = 0.2 +deess = 0.7 +presence = 1.0 +air = 0.8 +breath = 0.2 +# hpf_hz = 80.0 # high-pass corner frequency (podcast default 80) + +[radio] +mud = 0.9 +warmth = 1.0 +soothe = 0.7 +compress = 1.0 +tape = 0.55 +deess = 0.5 +presence = 0.7 +air = 0.5 +breath = 0.4 +# hpf_hz = 70.0 # high-pass corner frequency (radio default 70) |
