use nih_plug::prelude::*; use realfft::num_complex::Complex32; use realfft::{ComplexToReal, RealFftPlanner, RealToComplex}; use std::f32; use std::sync::Arc; /// The size of the windows we'll process at a time. const WINDOW_SIZE: usize = 64; /// The length of the filter's impulse response. const FILTER_WINDOW_SIZE: usize = 33; /// The length of the FFT window we will use to perform FFT convolution. This includes padding to /// prevent time domain aliasing as a result of cyclic convolution. const FFT_WINDOW_SIZE: usize = WINDOW_SIZE + FILTER_WINDOW_SIZE - 1; /// The gain compensation we need to apply for the STFT process. const GAIN_COMPENSATION: f32 = 1.0 / FFT_WINDOW_SIZE as f32; struct Stft { params: Arc, /// An adapter that performs most of the overlap-add algorithm for us. stft: util::StftHelper, /// The FFT of a simple low-pass FIR filter. lp_filter_spectrum: Vec, /// The algorithm for the FFT operation. r2c_plan: Arc>, /// The algorithm for the IFFT operation. c2r_plan: Arc>, /// The output of our real->complex FFT. complex_fft_buffer: Vec, } #[derive(Params)] struct StftParams {} impl Default for Stft { fn default() -> Self { let mut planner = RealFftPlanner::new(); let r2c_plan = planner.plan_fft_forward(FFT_WINDOW_SIZE); let c2r_plan = planner.plan_fft_inverse(FFT_WINDOW_SIZE); let mut real_fft_buffer = r2c_plan.make_input_vec(); let mut complex_fft_buffer = r2c_plan.make_output_vec(); // Build a super simple low-pass filter from one of the built in window functions let mut filter_window = util::window::hann(FILTER_WINDOW_SIZE); // And make sure to normalize this so convolution sums to 1 let filter_normalization_factor = filter_window.iter().sum::().recip(); for sample in &mut filter_window { *sample *= filter_normalization_factor; } real_fft_buffer[0..FILTER_WINDOW_SIZE].copy_from_slice(&filter_window); // RustFFT doesn't actually need a scratch buffer here, so we'll pass an empty buffer // instead r2c_plan .process_with_scratch(&mut real_fft_buffer, &mut complex_fft_buffer, &mut []) .unwrap(); Self { params: Arc::new(StftParams::default()), // We'll process the input in `WINDOW_SIZE` chunks, but our FFT window is slightly // larger to account for time domain aliasing so we'll need to add some padding ot each // block. stft: util::StftHelper::new(2, WINDOW_SIZE, FFT_WINDOW_SIZE - WINDOW_SIZE), lp_filter_spectrum: complex_fft_buffer.clone(), r2c_plan, c2r_plan, complex_fft_buffer, } } } #[allow(clippy::derivable_impls)] impl Default for StftParams { fn default() -> Self { Self {} } } impl Plugin for Stft { const NAME: &'static str = "STFT Example"; const VENDOR: &'static str = "Moist Plugins GmbH"; const URL: &'static str = "https://youtu.be/dQw4w9WgXcQ"; const EMAIL: &'static str = "info@example.com"; const VERSION: &'static str = "0.0.1"; const DEFAULT_NUM_INPUTS: u32 = 2; const DEFAULT_NUM_OUTPUTS: u32 = 2; const SAMPLE_ACCURATE_AUTOMATION: bool = true; fn params(&self) -> Arc { self.params.clone() } fn accepts_bus_config(&self, config: &BusConfig) -> bool { // We'll only do stereo for simplicity's sake config.num_input_channels == config.num_output_channels && config.num_input_channels == 2 } fn initialize( &mut self, _bus_config: &BusConfig, _buffer_config: &BufferConfig, context: &mut impl InitContext, ) -> bool { // The plugin's latency consists of the block size from the overlap-add procedure and half // of the filter kernel's size (since we're using a linear phase/symmetrical convolution // kernel) context.set_latency_samples(self.stft.latency_samples() + (FILTER_WINDOW_SIZE as u32 / 2)); true } fn reset(&mut self) { // Normally we'd also initialize the STFT helper for the correct channel count here, but we // only do stereo so that's not necessary. Setting the block size also zeroes out the // buffers. self.stft.set_block_size(WINDOW_SIZE); } fn process( &mut self, buffer: &mut Buffer, _aux: &mut AuxiliaryBuffers, _context: &mut impl ProcessContext, ) -> ProcessStatus { self.stft .process_overlap_add(buffer, 1, |_channel_idx, real_fft_buffer| { // Forward FFT, `real_fft_buffer` already is already padded with zeroes, and the // padding from the last iteration will have already been added back to the start of // the buffer self.r2c_plan .process_with_scratch(real_fft_buffer, &mut self.complex_fft_buffer, &mut []) .unwrap(); // As per the convolution theorem we can simply multiply these two buffers. We'll // also apply the gain compensation at this point. for (fft_bin, kernel_bin) in self .complex_fft_buffer .iter_mut() .zip(&self.lp_filter_spectrum) { *fft_bin *= *kernel_bin * GAIN_COMPENSATION; } // Inverse FFT back into the scratch buffer. This will be added to a ring buffer // which gets written back to the host at a one block delay. self.c2r_plan .process_with_scratch(&mut self.complex_fft_buffer, real_fft_buffer, &mut []) .unwrap(); }); ProcessStatus::Normal } } impl ClapPlugin for Stft { const CLAP_ID: &'static str = "com.moist-plugins-gmbh.stft"; const CLAP_DESCRIPTION: &'static str = "An example plugin using the STFT helper"; const CLAP_FEATURES: &'static [&'static str] = &["audio-effect", "stereo", "tool"]; const CLAP_MANUAL_URL: &'static str = Self::URL; const CLAP_SUPPORT_URL: &'static str = Self::URL; } impl Vst3Plugin for Stft { const VST3_CLASS_ID: [u8; 16] = *b"StftMoistestPlug"; const VST3_CATEGORIES: &'static str = "Fx|Tools"; } nih_export_clap!(Stft); nih_export_vst3!(Stft);