extern crate ferretro; extern crate ffmpeg4 as ffmpeg; use std::collections::VecDeque; use std::env; use std::path::{Path, PathBuf}; use std::pin::Pin; use ferretro::retro; use ferretro::retro::ffi::{PixelFormat, GameGeometry, SystemAvInfo}; use ferretro::retro::wrapper::LibretroWrapper; use ffmpeg::{codec, filter, format, frame, media, ChannelLayout}; use ffmpeg::{rescale, Rescale}; struct MyEmulator { retro: retro::wrapper::LibretroWrapper, audio_buf: Vec<(i16, i16)>, video_pixfmt: format::Pixel, video_frames: VecDeque, video_encoder: ffmpeg::encoder::Video, audio_encoder: ffmpeg::encoder::Audio, sys_path: Option, } impl MyEmulator { pub fn new(core_path: impl AsRef, sys_path: &Option>, ffstream: format::stream::StreamMut) -> Pin> { let lib = libloading::Library::new(core_path).unwrap(); let raw_retro = retro::loading::LibretroApi::from_library(lib).unwrap(); let retro = retro::wrapper::LibretroWrapper::from(raw_retro); let mut av_info = retro.get_system_av_info(); let video_pixfmt = format::Pixel::RGB555; let emu = MyEmulator { retro, audio_buf: Default::default(), video_pixfmt, video_frames: Default::default(), video_encoder, audio_encoder, sys_path: sys_path.map(|x| x.as_ref().to_path_buf()), }; let mut pin_emu = Box::pin(emu); retro::wrapper::set_handler(pin_emu.as_mut()); pin_emu.retro.init(); pin_emu } pub fn run(&mut self) { self.retro.run(); let vframe = self.video_frames.pop_front().unwrap(); let mut aframe = frame::Audio::new( format::Sample::I16(format::sample::Type::Packed), self.audio_buf.len(), ChannelLayout::STEREO ); let aplane: &mut [(i16, i16)] = aframe.plane_mut(0); aplane.copy_from_slice(self.audio_buf.as_ref()); self.audio_buf.clear(); let mut out = ffmpeg::Packet::empty(); self.video_encoder.encode(&vframe, &mut out); self.audio_encoder.encode(&aframe, &mut out); } } impl retro::wrapper::Handler for MyEmulator { fn libretro_core(&mut self) -> &mut LibretroWrapper { &mut self.retro } fn video_refresh(&mut self, data: &[u8], width: u32, height: u32, pitch: u32) { let mut vframe = frame::Video::new(self.video_pixfmt, width, height); let stride = vframe.stride(0); let pitch = pitch as usize; let vplane = vframe.data_mut(0); if data.len() == vplane.len() && pitch == stride { vplane.copy_from_slice(&data); } else { for y in 0..(height as usize) { let ffbegin = y * stride; let lrbegin = y * pitch; let min = usize::min(stride, pitch); let x = vplane[ffbegin..(ffbegin + min)].copy_from_slice( &data[lrbegin..(lrbegin + min)] ); } } self.video_frames.push_back(vframe); } fn audio_sample(&mut self, left: i16, right: i16) { self.audio_buf.push((left, right)); } fn audio_sample_batch(&mut self, stereo_pcm: &[i16]) -> usize { let left_iter = stereo_pcm.iter().step_by(2).cloned(); let right_iter = stereo_pcm.iter().skip(1).step_by(2).cloned(); self.audio_buf.extend(Iterator::zip(left_iter, right_iter)); stereo_pcm.len() } fn set_pixel_format(&mut self, format: PixelFormat) -> bool { self.video_pixfmt = match format { PixelFormat::ARGB1555 => format::Pixel::RGB555, PixelFormat::ARGB8888 => format::Pixel::RGB32, PixelFormat::RGB565 => format::Pixel::RGB565, }; true } fn set_system_av_info(&mut self, system_av_info: SystemAvInfo) -> bool { self.video_encoder.set_frame_rate(system_av_info.timing.fps.into()); self.audio_encoder.set_rate(system_av_info.timing.sample_rate.round() as i32); self.set_geometry(system_av_info.geometry); true } fn set_geometry(&mut self, geometry: GameGeometry) -> bool { self.video_encoder.set_width(geometry.base_width); self.video_encoder.set_height(geometry.base_height); self.video_encoder.set_aspect_ratio(geometry.aspect_ratio as f64); true } } fn filter( spec: &str, decoder: &codec::decoder::Audio, encoder: &codec::encoder::Audio, ) -> Result { let mut filter = filter::Graph::new(); let args = format!( "time_base={}:sample_rate={}:sample_fmt={}:channel_layout=0x{:x}", decoder.time_base(), decoder.rate(), decoder.format().name(), decoder.channel_layout().bits() ); filter.add(&filter::find("abuffer").unwrap(), "in", &args)?; filter.add(&filter::find("abuffersink").unwrap(), "out", "")?; { let mut out = filter.get("out").unwrap(); out.set_sample_format(encoder.format()); out.set_channel_layout(encoder.channel_layout()); out.set_sample_rate(encoder.rate()); } filter.output("in", 0)?.input("out", 0)?.parse(spec)?; filter.validate()?; println!("{}", filter.dump()); if let Some(codec) = encoder.codec() { if !codec .capabilities() .contains(ffmpeg::codec::capabilities::Capabilities::VARIABLE_FRAME_SIZE) { filter .get("out") .unwrap() .sink() .set_frame_size(encoder.frame_size()); } } Ok(filter) } struct Transcoder { stream: usize, filter: filter::Graph, decoder: codec::decoder::Audio, encoder: codec::encoder::Audio, } fn transcoder>( ictx: &mut format::context::Input, octx: &mut format::context::Output, path: &P, filter_spec: &str, ) -> Result { let input = ictx .streams() .best(media::Type::Audio) .expect("could not find best audio stream"); let mut decoder = input.codec().decoder().audio()?; let codec = ffmpeg::encoder::find(octx.format().codec(path, media::Type::Audio)) .expect("failed to find encoder") .audio()?; let global = octx .format() .flags() .contains(ffmpeg::format::flag::Flags::GLOBAL_HEADER); decoder.set_parameters(input.parameters())?; let mut output = octx.add_stream(codec)?; let mut encoder = output.codec().encoder().audio()?; let channel_layout = codec .channel_layouts() .map(|cls| cls.best(decoder.channel_layout().channels())) .unwrap_or(ffmpeg::channel_layout::ChannelLayout::STEREO); if global { encoder.set_flags(ffmpeg::codec::flag::Flags::GLOBAL_HEADER); } encoder.set_rate(decoder.rate() as i32); encoder.set_channel_layout(channel_layout); encoder.set_channels(channel_layout.channels()); encoder.set_format( codec .formats() .expect("unknown supported formats") .next() .unwrap(), ); encoder.set_bit_rate(decoder.bit_rate()); encoder.set_max_bit_rate(decoder.max_bit_rate()); encoder.set_time_base((1, decoder.rate() as i32)); output.set_time_base((1, decoder.rate() as i32)); let encoder = encoder.open_as(codec)?; output.set_parameters(&encoder); let filter = filter(filter_spec, &decoder, &encoder)?; Ok(Transcoder { stream: input.index(), filter: filter, decoder: decoder, encoder: encoder, }) } // Transcode the `best` audio stream of the input file into a the output file while applying a // given filter. If no filter was specified the stream gets copied (`anull` filter). // // Example 1: Transcode *.mp3 file to *.wmv while speeding it up // transcode-audio in.mp3 out.wmv "atempo=1.2" // // Example 2: Overlay an audio file // transcode-audio in.mp3 out.mp3 "amovie=overlay.mp3 [ov]; [in][ov] amerge [out]" // // Example 3: Seek to a specified position (in seconds) // transcode-audio in.mp3 out.mp3 anull 30 fn main() { ffmpeg::init().unwrap(); let input = env::args().nth(1).expect("missing input"); let output = env::args().nth(2).expect("missing output"); let filter = env::args().nth(3).unwrap_or_else(|| "anull".to_owned()); let seek = env::args().nth(4).and_then(|s| s.parse::().ok()); let mut ictx = format::input(&input).unwrap(); let mut octx = format::output(&output).unwrap(); let mut transcoder = transcoder(&mut ictx, &mut octx, &output, &filter).unwrap(); if let Some(position) = seek { // If the position was given in seconds, rescale it to ffmpegs base timebase. let position = position.rescale((1, 1), rescale::TIME_BASE); // If this seek was embedded in the transcoding loop, a call of `flush()` // for every opened buffer after the successful seek would be advisable. ictx.seek(position, ..position).unwrap(); } octx.set_metadata(ictx.metadata().to_owned()); octx.write_header().unwrap(); let in_time_base = transcoder.decoder.time_base(); let out_time_base = octx.stream(0).unwrap().time_base(); let mut decoded = frame::Audio::empty(); let mut encoded = ffmpeg::Packet::empty(); for (stream, mut packet) in ictx.packets() { if stream.index() == transcoder.stream { packet.rescale_ts(stream.time_base(), in_time_base); if let Ok(true) = transcoder.decoder.decode(&packet, &mut decoded) { let timestamp = decoded.timestamp(); decoded.set_pts(timestamp); transcoder .filter .get("in") .unwrap() .source() .add(&decoded) .unwrap(); while let Ok(..) = transcoder .filter .get("out") .unwrap() .sink() .frame(&mut decoded) { if let Ok(true) = transcoder.encoder.encode(&decoded, &mut encoded) { encoded.set_stream(0); encoded.rescale_ts(in_time_base, out_time_base); encoded.write_interleaved(&mut octx).unwrap(); } } } } } transcoder .filter .get("in") .unwrap() .source() .flush() .unwrap(); while let Ok(..) = transcoder .filter .get("out") .unwrap() .sink() .frame(&mut decoded) { if let Ok(true) = transcoder.encoder.encode(&decoded, &mut encoded) { encoded.set_stream(0); encoded.rescale_ts(in_time_base, out_time_base); encoded.write_interleaved(&mut octx).unwrap(); } } if let Ok(true) = transcoder.encoder.flush(&mut encoded) { encoded.set_stream(0); encoded.rescale_ts(in_time_base, out_time_base); encoded.write_interleaved(&mut octx).unwrap(); } octx.write_trailer().unwrap(); }