You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

489 lines
16 KiB
Rust

use crate::av::avio::{AVIOWriter, AVIO};
use crate::av::decoder::Decoder;
use crate::av::dictionary::Dictionary;
use crate::av::encoder::Encoder;
use crate::av::format::Format;
use crate::av::mov_avc::{annex_b_to_avc, AvcDecoderConfigurationRecord, NalBuffer, NalUnitType};
use crate::av::packet::Packet;
use crate::av::resampler::Resampler;
use crate::av::scaler::Scaler;
use crate::av::stream::Stream;
use crate::av::xcoder::XCoder;
use crate::av::Rational;
use crate::utils::SortedFrameBuffer;
use ffmpeg_sys_next::AVMediaType::{AVMEDIA_TYPE_AUDIO, AVMEDIA_TYPE_VIDEO};
use ffmpeg_sys_next::AVPictureType::AV_PICTURE_TYPE_I;
use ffmpeg_sys_next::AVPixelFormat::AV_PIX_FMT_YUV420P;
use ffmpeg_sys_next::{avio_wb32, AVIOContext, AVPixelFormat, SWS_BILINEAR};
use num_rational::Ratio;
use std::fs::{DirBuilder, File};
use std::io::Write;
use std::option::Option::Some;
pub struct Transcoder {
input: Format<()>,
input_video: Stream,
input_frame_rate: Ratio<i32>,
video_output: Format<Buffer>,
output_path: String,
output_video: Stream,
video_frame_buffer: SortedFrameBuffer,
video_encoder: Encoder,
video_decoder: Decoder,
frame_scaler: Scaler,
last_pts: i64,
current_pts: i64,
video_segment: u32,
extra_data: Option<Vec<u8>>,
audio: Option<AudioTranscoder>,
}
pub struct AudioTranscoder {
encoder: Encoder,
decoder: Decoder,
frame_buffer: SortedFrameBuffer,
output_stream: Stream,
input_stream: Stream,
format: Format<Buffer>,
output_path: String,
resampler: Resampler,
segment: u32,
last_pts: i64,
current_pts: i64,
}
impl AudioTranscoder {
fn process_packet(&mut self, packet: Packet) -> Result<(), String> {
self.decoder
.send_packet(&packet)
.map_err(|err| format!("Failed sending audio packet: {}", err))?;
while let Some(frame) = self.decoder.read_frame()? {
self.frame_buffer.insert_frame(frame);
self.encode()?;
}
Ok(())
}
fn start_segment(&self, force: bool) {
if self.segment == 0 && !force {
return;
}
unsafe {
let pb = (*self.format.as_mut_ptr()).pb;
avio_wb32(pb, 24);
ffio_wfourcc(pb, b"styp");
ffio_wfourcc(pb, b"msdh");
avio_wb32(pb, 0); /* minor */
ffio_wfourcc(pb, b"msdh");
ffio_wfourcc(pb, b"msix");
}
}
fn write_header(&mut self) -> Result<(), String> {
File::create(format!("{}/audio-init.mp4", &self.output_path))
.unwrap()
.write_all(&mut self.format.avio_inner_mut().unwrap().buffer())
.map_err(|_| "Failed to write audio init segment".to_string())?;
self.start_segment(true);
self.format.write_packet_null()?;
Ok(())
}
fn write_segment(&mut self) -> Result<(), String> {
if self.segment == 0 {
self.write_header()?;
}
if let Some(avio) = self.format.avio.as_mut() {
avio.flush()
}
let mut segment = self.format.avio_inner_mut().unwrap().buffer();
File::create(format!(
"{}/audio-segment-{:0>5}.m4s",
self.output_path, self.segment
))
.unwrap()
.write_all(&mut segment)
.map_err(|_| format!("Failed to write audio segment {}", self.segment))?;
self.start_segment(false);
Ok(())
}
fn encode(&mut self) -> Result<(), String> {
while let Some(frame) = self.frame_buffer.pop_first() {
let pts_passed = frame.pts() - self.last_pts;
if (pts_passed + self.current_pts) > (5 * self.input_stream.time_base().den()) as i64 {
self.format.write_packet_null()?;
println!(
"Next audio segment (current: {}, pts: {}/{})!",
self.segment, self.last_pts, self.current_pts,
);
self.write_segment()?;
self.segment += 1;
self.current_pts =
(pts_passed + self.current_pts) % self.input_stream.time_base().den() as i64;
} else {
self.current_pts += pts_passed
}
self.last_pts = frame.pts();
self.encoder.send_audio_frame(&frame)?;
while let Some(new_packet) = self.encoder.read_packet()? {
new_packet.set_pts(frame.pts());
new_packet.set_dts(new_packet.pts());
new_packet.set_duration(pts_passed);
new_packet.rescale(
self.input_stream.time_base(),
self.output_stream.time_base(),
);
new_packet.set_stream(0);
self.format.write_packet(&new_packet)?;
}
}
Ok(())
}
}
impl Transcoder {
pub fn create(
input_path: &str,
output_path: &str,
video_index: Option<i32>,
audio_index: Option<i32>,
) -> Result<Transcoder, String> {
let input = Format::open(input_path, None).expect("Failed to open media");
let output_path = output_path.to_string();
let input_video = input
.stream(AVMEDIA_TYPE_VIDEO, video_index)
.ok_or("Failed to find video stream".to_string())?;
let input_audio = input.stream(AVMEDIA_TYPE_AUDIO, audio_index);
if audio_index.is_some() && input_audio.is_none() {
return Err("Failed to find audio stream".to_string());
}
let input_frame_rate = input_video.avg_frame_rate(&input).unwrap().to_num();
let mut options = Dictionary::new();
options.set("movflags", "+dash+delay_moov+skip_trailer+frag_custom");
let video_avio = AVIO::writer(Buffer::default());
let video_output = Format::output_avio(video_avio, "mp4")?;
let output_video = video_output.new_stream("h264")?;
output_video.params().copy_from(&input_video.params())?;
output_video.set_sample_aspect_ratio(input_video.sample_aspect_ratio());
video_output.set_flags(input.flags());
let audio = if let Some(input_stream) = input_audio {
let audio_avio = AVIO::writer(Buffer::default());
let format = Format::output_avio(audio_avio, "mp4")?;
let output_stream = format.new_stream("aac")?;
format.set_flags(input.flags());
let decoder = input_stream.decoder(None).ok_or(format!(
"Couldn't find encoder for input audio with codec: {:?}",
input_stream.codec()
))?;
decoder.configure(&input_stream)?;
decoder.open()?;
let encoder = output_stream
.encoder(None)
.ok_or("Couldn't find encoder for AAC".to_string())?;
encoder.set_channels(decoder.channels());
encoder.set_channel_layout(decoder.channel_layout());
encoder.set_sample_format(
encoder
.sample_formats()
.first()
.copied()
.ok_or("No sample formats found for AAC")?,
);
encoder.set_sample_rate(decoder.sample_rate());
encoder.set_time_base(input_stream.time_base());
encoder.set_time_base(Ratio::new(1, decoder.sample_rate()));
output_stream.set_time_base(encoder.time_base());
encoder.set_frame_size(decoder.frame_size() * 4);
encoder.open()?;
encoder.configure(&output_stream)?;
println!(
"audio decoder[{}] -> audio encoder[{}]",
decoder.name(),
encoder.name()
);
let resampler = Resampler::from_coder(&decoder, &encoder);
Some(AudioTranscoder {
frame_buffer: SortedFrameBuffer::new(),
encoder,
decoder,
format,
output_stream,
input_stream,
output_path: output_path.to_string(),
current_pts: 0,
last_pts: 0,
segment: 0,
resampler,
})
} else {
None
};
let video_decoder = input_video.decoder_select(None, |d| {
d.configure(&input_video)?;
d.open()?;
Ok(())
})?;
let video_encoder = output_video.encoder_select(None, |encoder| {
encoder.set_pixel_format(AVPixelFormat::AV_PIX_FMT_YUV420P);
encoder.set_color_range(video_decoder.color_range());
encoder.set_color_primaries(video_decoder.color_primaries());
encoder.set_color_trc(video_decoder.color_trc());
encoder.set_height(video_decoder.height());
encoder.set_width(video_decoder.width());
encoder.set_time_base(input_frame_rate.invert());
output_video.set_time_base(encoder.time_base());
encoder.open()?;
encoder.configure(&output_video)?;
Ok(())
})?;
println!(
"video decoder[{}] -> video encoder[{}]",
video_decoder.name(),
video_encoder.name()
);
let param = output_video.params();
param.set_height(video_decoder.height());
param.set_width(video_decoder.width());
param.set_pixel_format(AV_PIX_FMT_YUV420P);
video_output.init_output(options)?;
let frame_scaler = Scaler::from_coder(&video_decoder, &video_encoder, SWS_BILINEAR);
let video_frame_buffer = SortedFrameBuffer::new();
Ok(Transcoder {
input,
input_video,
input_frame_rate,
video_output,
output_video,
audio,
output_path,
video_encoder,
video_decoder,
frame_scaler,
video_frame_buffer,
last_pts: 0,
current_pts: 0,
video_segment: 0,
extra_data: None,
})
}
pub fn transcode(&mut self) -> Result<(), String> {
DirBuilder::new()
.recursive(true)
.create(self.output_path.to_string())
.map_err(|_| "Failed to create target directory".to_string())?;
self.video_output.write_header(None).unwrap();
let input_video_index = self.input_video.index();
let input_audio_index = self.audio.as_ref().map(|ia| ia.input_stream.index());
while let Some(packet) = self.input.next_packet() {
if input_video_index == packet.stream() {
self.video_process_packet(packet)?;
} else if input_audio_index == Some(packet.stream()) {
// Safe assumption
if let Some(ref mut audio) = &mut self.audio {
audio.process_packet(packet)?;
}
}
}
self.encode_video()?;
self.video_output.write_trailer()?;
self.video_write_segment()?;
Ok(())
}
fn video_process_packet(&mut self, packet: Packet) -> Result<(), String> {
self.video_decoder.send_packet(&packet)?;
while let Some(frame) = self.video_decoder.read_frame()? {
let pts = frame.pts();
let frame = self
.frame_scaler
.scale(frame)
.expect("Failed to scale video frame");
frame.set_pts(pts);
self.video_frame_buffer.insert_frame(frame);
self.encode_video()?;
}
Ok(())
}
fn video_write_header(&mut self) -> Result<(), String> {
File::create(format!("{}/video-init.mp4", &self.output_path))
.unwrap()
.write_all(&mut self.video_output.avio_inner_mut().unwrap().buffer())
.map_err(|_| "Failed to write video init segment".to_string())?;
self.start_segment(true);
self.video_output.write_packet_null()?;
Ok(())
}
fn video_write_segment(&mut self) -> Result<(), String> {
if self.video_segment == 0 {
self.video_write_header()?;
}
if let Some(avio) = self.video_output.avio.as_mut() {
avio.flush()
}
let segment = self.video_output.avio_inner_mut().unwrap().buffer();
let mut segment = annex_b_to_avc(segment, 4)?;
File::create(format!(
"{}/video-segment-{:0>5}.m4s",
self.output_path, self.video_segment
))
.unwrap()
.write_all(&mut segment)
.map_err(|_| format!("Failed to write video segment {}", self.video_segment))?;
self.start_segment(false);
Ok(())
}
fn encode_video(&mut self) -> Result<(), String> {
while let Some(frame) = self.video_frame_buffer.pop_first() {
let pts_passed = frame.pts() - self.last_pts;
if (pts_passed + self.current_pts) > (5 * self.input_frame_rate.den()) as i64 {
self.video_output.write_packet_null()?;
println!(
"Next video segment (current: {}, pts: {}/{})!",
self.video_segment, self.last_pts, self.current_pts,
);
self.video_write_segment()?;
self.video_segment += 1;
self.current_pts =
(pts_passed + self.current_pts) % self.input_frame_rate.den() as i64;
frame.set_pict_type(AV_PICTURE_TYPE_I);
} else {
self.current_pts += pts_passed
}
self.last_pts = frame.pts();
self.video_encoder.send_video_frame(&frame)?;
while let Some(new_packet) = self.video_encoder.read_packet()? {
if self.extra_data.is_none() {
let items = NalBuffer::from_stream(new_packet.data())?;
let mut sps = None;
let mut pss = None;
for item in items {
if item.nal_unit.nal_unit_type == NalUnitType::PictureParameterSet {
pss = Some(item.to_vec());
} else if item.nal_unit.nal_unit_type == NalUnitType::SequenceParameterSet {
sps = Some(item.to_vec());
}
}
if let Some(sps) = sps {
if let Some(pss) = pss {
let config =
AvcDecoderConfigurationRecord::from_parameter_sets(sps, pss);
self.extra_data = Some(config.to_bytes()?);
}
}
if let Some(extra_data) = &self.extra_data {
self.output_video.params().set_extra_data(extra_data);
self.output_video
.params()
.set_sample_aspect_ratio(Ratio::new(1, 1));
}
}
new_packet.set_pts(frame.pts());
new_packet.set_dts(new_packet.pts());
new_packet.set_duration(pts_passed);
new_packet.rescale(self.input_video.time_base(), self.output_video.time_base());
new_packet.set_stream(0);
self.video_output.write_packet(&new_packet)?;
}
}
Ok(())
}
fn start_segment(&self, force: bool) {
if self.video_segment == 0 && !force {
return;
}
unsafe {
let pb = (*self.video_output.as_mut_ptr()).pb;
avio_wb32(pb, 24);
ffio_wfourcc(pb, b"styp");
ffio_wfourcc(pb, b"msdh");
avio_wb32(pb, 0); /* minor */
ffio_wfourcc(pb, b"msdh");
ffio_wfourcc(pb, b"msix");
}
}
}
#[derive(Clone, Default, Debug)]
struct Buffer {
buffer: Vec<u8>,
}
impl Buffer {
fn buffer(&mut self) -> Vec<u8> {
std::mem::replace(&mut self.buffer, vec![])
}
}
impl AVIOWriter for Buffer {
fn write(&mut self, mut buffer: Vec<u8>) {
self.buffer.append(&mut buffer)
}
}
#[inline]
fn ffio_wfourcc(pb: *mut AVIOContext, tag: &[u8]) {
let mut nr: u32 = 0;
nr |= tag[3] as u32;
nr |= (tag[2] as u32) << 8;
nr |= (tag[1] as u32) << 16;
nr |= (tag[0] as u32) << 24;
unsafe { avio_wb32(pb, nr) }
}