Use rust-native chromaprint port instead of ffmpeg

This commit is contained in:
bytedream 2024-04-29 03:03:25 +02:00
parent d3a227f364
commit 7837295f9b
5 changed files with 79 additions and 96 deletions

96
Cargo.lock generated
View file

@ -179,18 +179,6 @@ version = "3.16.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "79296716171880943b8470b5f8d03aa55eb2e645a4874bdbb28adb49162e012c"
[[package]]
name = "bytemuck"
version = "1.15.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5d6d68c57235a3a081186990eca2867354726650f42f7516ca50c28d6281fd15"
[[package]]
name = "byteorder"
version = "1.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
[[package]]
name = "bytes"
version = "1.6.0"
@ -381,8 +369,6 @@ dependencies = [
"fs2",
"futures-util",
"http",
"image",
"image_hasher",
"indicatif",
"lazy_static",
"log",
@ -391,6 +377,7 @@ dependencies = [
"regex",
"reqwest",
"rustls-native-certs",
"rusty-chromaprint",
"serde",
"serde_json",
"serde_plain",
@ -951,32 +938,6 @@ dependencies = [
"unicode-normalization",
]
[[package]]
name = "image"
version = "0.25.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fd54d660e773627692c524beaad361aca785a4f9f5730ce91f42aabe5bce3d11"
dependencies = [
"bytemuck",
"byteorder",
"num-traits",
"zune-core",
"zune-jpeg",
]
[[package]]
name = "image_hasher"
version = "2.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9481465fe767d92494987319b0b447a5829edf57f09c52bf8639396abaaeaf78"
dependencies = [
"base64 0.22.0",
"image",
"rustdct",
"serde",
"transpose",
]
[[package]]
name = "indexmap"
version = "1.9.3"
@ -1417,6 +1378,15 @@ dependencies = [
"proc-macro2",
]
[[package]]
name = "realfft"
version = "3.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "953d9f7e5cdd80963547b456251296efc2626ed4e3cbf36c869d9564e0220571"
dependencies = [
"rustfft",
]
[[package]]
name = "redox_users"
version = "0.4.5"
@ -1531,21 +1501,24 @@ version = "0.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b833d8d034ea094b1ea68aa6d5c740e0d04bad9d16568d08ba6f76823a114316"
[[package]]
name = "rubato"
version = "0.14.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e6dd52e80cfc21894deadf554a5673002938ae4625f7a283e536f9cf7c17b0d5"
dependencies = [
"num-complex",
"num-integer",
"num-traits",
"realfft",
]
[[package]]
name = "rustc-demangle"
version = "0.1.23"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d626bb9dae77e28219937af045c257c28bfd3f69333c512553507f5f9798cb76"
[[package]]
name = "rustdct"
version = "0.7.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8b61555105d6a9bf98797c063c362a1d24ed8ab0431655e38f1cf51e52089551"
dependencies = [
"rustfft",
]
[[package]]
name = "rustfft"
version = "6.2.0"
@ -1628,6 +1601,16 @@ dependencies = [
"untrusted",
]
[[package]]
name = "rusty-chromaprint"
version = "0.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1755646867c36ecb391776deaa0b557a76d3badf20c142de7282630c34b20440"
dependencies = [
"rubato",
"rustfft",
]
[[package]]
name = "ryu"
version = "1.0.17"
@ -2501,18 +2484,3 @@ name = "zeroize"
version = "1.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "63381fa6624bf92130a6b87c0d07380116f80b565c42cf0d754136f0238359ef"
[[package]]
name = "zune-core"
version = "0.4.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3f423a2c17029964870cfaabb1f13dfab7d092a62a29a89264f4d36990ca414a"
[[package]]
name = "zune-jpeg"
version = "0.4.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ec866b44a2a1fd6133d363f073ca1b179f438f99e7e5bfb1e33f7181facfe448"
dependencies = [
"zune-core",
]

View file

@ -24,14 +24,13 @@ derive_setters = "0.1"
futures-util = { version = "0.3", features = ["io"] }
fs2 = "0.4"
http = "1.1"
image = { version = "0.25", features = ["jpeg"], default-features = false }
image_hasher = "2.0"
indicatif = "0.17"
lazy_static = "1.4"
log = { version = "0.4", features = ["std"] }
num_cpus = "1.16"
regex = "1.10"
reqwest = { version = "0.12", features = ["socks", "stream"] }
rusty-chromaprint = "0.2"
serde = "1.0"
serde_json = "1.0"
serde_plain = "1.0"

View file

@ -281,6 +281,7 @@ impl Downloader {
format_id: i,
path,
locale: locale.clone(),
sample_rate: stream_data.sampling_rate().unwrap(),
video_idx: i,
})
}

View file

@ -1,6 +1,9 @@
use std::io::Read;
use std::process::Stdio;
use std::{
cmp,
collections::{HashMap, HashSet},
mem,
ops::Not,
path::Path,
process::Command,
@ -12,6 +15,7 @@ use log::debug;
use tempfile::TempPath;
use anyhow::{bail, Result};
use rusty_chromaprint::{Configuration, Fingerprinter};
use super::fmt::format_time_delta;
@ -19,6 +23,7 @@ pub struct SyncAudio {
pub format_id: usize,
pub path: TempPath,
pub locale: Locale,
pub sample_rate: u32,
pub video_idx: usize,
}
@ -43,11 +48,12 @@ pub fn sync_audios(
continue;
}
formats.insert(audio.format_id);
sync_audios.push((audio.format_id, &audio.path));
sync_audios.push((audio.format_id, &audio.path, audio.sample_rate));
chromaprints.insert(
audio.format_id,
generate_chromaprint(
&audio.path,
audio.sample_rate,
&TimeDelta::zero(),
&TimeDelta::zero(),
&TimeDelta::zero(),
@ -112,6 +118,7 @@ pub fn sync_audios(
for sync_audio in &sync_audios {
let chromaprint = generate_chromaprint(
sync_audio.1,
sync_audio.2,
&start,
&end,
initial_offsets.get(&sync_audio.0).unwrap(),
@ -127,7 +134,7 @@ pub fn sync_audios(
);
chromaprints.insert(
base_audio.0,
generate_chromaprint(base_audio.1, &start, &end, &base_offset)?,
generate_chromaprint(base_audio.1, base_audio.2, &start, &end, &base_offset)?,
);
for audio in &sync_audios {
let initial_offset = initial_offsets.get(&audio.0).copied().unwrap();
@ -207,6 +214,7 @@ fn find_offset(
fn generate_chromaprint(
input_file: &Path,
sample_rate: u32,
start: &TimeDelta,
end: &TimeDelta,
offset: &TimeDelta,
@ -218,6 +226,9 @@ fn generate_chromaprint(
offset_argument = offset;
};
let mut printer = Fingerprinter::new(&Configuration::preset_test1());
printer.start(sample_rate, 2)?;
let mut command = Command::new("ffmpeg");
command
.arg("-hide_banner")
@ -232,30 +243,42 @@ fn generate_chromaprint(
.args(["-itsoffset", format_time_delta(offset_argument).as_str()])
.args(["-i", input_file.to_string_lossy().to_string().as_str()])
.args(["-ac", "2"])
.args(["-f", "chromaprint"])
.args(["-fp_format", "raw"])
.args([
"-f",
if cfg!(target_endian = "big") {
"s16be"
} else {
"s16le"
},
])
.arg("-");
let extract_output = command.output()?;
let mut handle = command
.stdout(Stdio::piped())
.stderr(Stdio::piped())
.spawn()?;
if !extract_output.status.success() {
bail!(
"{}",
String::from_utf8_lossy(extract_output.stderr.as_slice())
);
// the stdout is read in chunks because keeping all the raw audio data in memory would take up
// a significant amount of space
let mut stdout = handle.stdout.take().unwrap();
let mut buf: [u8; 32_000] = [0; 32_000];
while handle.try_wait()?.is_none() {
loop {
let read_bytes = stdout.read(&mut buf)?;
if read_bytes == 0 {
break;
}
let raw_chromaprint = extract_output.stdout.as_slice();
let length = raw_chromaprint.len();
if length % 4 != 0 {
bail!("chromaprint bytes should be a multiple of 4");
let data: [i16; 16_000] = unsafe { mem::transmute(buf) };
printer.consume(&data[0..(read_bytes / 2)])
}
let mut chromaprint = Vec::with_capacity(length / 4);
for i in 0..length / 4 {
chromaprint.push(as_u32_le(
raw_chromaprint[i * 4..i * 4 + 4].try_into().unwrap(),
));
}
Ok(chromaprint)
if !handle.wait()?.success() {
bail!("{}", std::io::read_to_string(handle.stderr.unwrap())?)
}
printer.finish();
return Ok(printer.fingerprint().into());
}
fn compare_chromaprints(
@ -407,11 +430,3 @@ fn timestamps_to_ranges(mut timestamps: Vec<f64>) -> Option<Vec<TimeRange>> {
None
}
}
fn as_u32_le(array: &[u8; 4]) -> u32 {
#![allow(arithmetic_overflow)]
(array[0] as u32)
| ((array[1] as u32) << 8)
| ((array[2] as u32) << 16)
| ((array[3] as u32) << 24)
}