inferencelayer 0.2.9

//! `chatterbox-wmcheck` — is this audio file one of ours?
//!
//! A watermark nobody can check is decoration, so the detector ships as a tool, not just a library
//! function. Reads a 24 kHz mono PCM16 WAV and prints the provenance score.
//!
//! Usage:
//!   cargo run --release --features "cli encoder-cpu" --bin chatterbox-wmcheck -- FILE.wav
//!
//! Exit status: 0 = mark present, 1 = absent, 2 = unreadable. So it composes in a shell:
//!   chatterbox-wmcheck out.wav && echo "generated by us"
//!
//! Scope, so results are not over-read: this detects THIS engine's mark (see `watermark_detect`).
//! It is NOT Resemble's Perth and shares nothing with it, so a "no" means "not marked by this
//! engine", never "not synthetic". The mark also lives above 8 kHz, so a 16 kHz or otherwise
//! lowpassed copy will read as unmarked even though the original was marked.

use anyhow::{Context, Result, bail};
use inferencelayer::chatterbox::{WATERMARK_ALPHA, WATERMARK_KEY, watermark_detect};

fn main() -> Result<()> {
    let Some(path) = std::env::args().nth(1) else {
        eprintln!("usage: chatterbox-wmcheck FILE.wav");
        std::process::exit(2);
    };
    let (samples, sr) = match read_wav(&path) {
        Ok(v) => v,
        Err(e) => {
            eprintln!("cannot read {path}: {e}");
            std::process::exit(2);
        }
    };
    let score = watermark_detect(&samples, WATERMARK_KEY);
    let thresh = WATERMARK_ALPHA / 2.0;
    let present = score > thresh;
    println!(
        "{path}: {:.2} s @ {sr} Hz | score {score:.4} (threshold {thresh:.4}) → {}",
        samples.len() as f32 / sr as f32,
        if present { "MARKED by this engine" } else { "no mark found" }
    );
    if sr != 24_000 {
        println!(
            "  note: {sr} Hz input — the mark sits above 8 kHz, so anything resampled below \
             16 kHz will read as unmarked regardless of its origin"
        );
    }
    std::process::exit(if present { 0 } else { 1 });
}

/// Minimal RIFF/PCM16 (or float32) WAV reader → (mono f32, sample rate).
fn read_wav(path: &str) -> Result<(Vec<f32>, u32)> {
    let b = std::fs::read(path).with_context(|| format!("read {path}"))?;
    if b.len() < 44 || &b[0..4] != b"RIFF" || &b[8..12] != b"WAVE" {
        bail!("not a RIFF/WAVE file");
    }
    let (mut pos, mut fmt) = (12usize, None);
    while pos + 8 <= b.len() {
        let id = &b[pos..pos + 4];
        let sz = u32::from_le_bytes([b[pos + 4], b[pos + 5], b[pos + 6], b[pos + 7]]) as usize;
        let body = pos + 8;
        if id == b"fmt " && body + 16 <= b.len() {
            let format = u16::from_le_bytes([b[body], b[body + 1]]);
            let ch = u16::from_le_bytes([b[body + 2], b[body + 3]]) as usize;
            let sr = u32::from_le_bytes([b[body + 4], b[body + 5], b[body + 6], b[body + 7]]);
            let bits = u16::from_le_bytes([b[body + 14], b[body + 15]]);
            fmt = Some((format, ch.max(1), sr, bits));
        } else if id == b"data" {
            let (format, ch, sr, bits) = fmt.context("data chunk before fmt chunk")?;
            let end = (body + sz).min(b.len());
            let raw = &b[body..end];
            let samples: Vec<f32> = match (format, bits) {
                (1, 16) => raw
                    .chunks_exact(2)
                    .step_by(ch) // first channel only
                    .map(|c| i16::from_le_bytes([c[0], c[1]]) as f32 / 32768.0)
                    .collect(),
                (3, 32) => raw
                    .chunks_exact(4)
                    .step_by(ch)
                    .map(|c| f32::from_le_bytes([c[0], c[1], c[2], c[3]]))
                    .collect(),
                _ => bail!("unsupported WAV encoding (format {format}, {bits} bit)"),
            };
            if samples.is_empty() {
                bail!("no samples");
            }
            return Ok((samples, sr));
        }
        pos = body + sz + (sz & 1);
    }
    bail!("no data chunk")
}