feat(*): added support for copying audio/subtitle streams and pause/abort (#1179)

* feat: added Makefile target for debian
* fix: fixed Dockerfile installing the wrong package
* feat: added hwaccel for encoder and decoder
* feat: added benchmark mode
* feat: removed hard-coded keyframe info
* chore: cleaned up headers and organized code
* style: cleaned up headers and includes
* feat: added a progress bar for CLI
* feat: removed atomicity requirements on processed frames
* feat: added pause and abort for CLI
* chore: updated default preset and crf settings
* feat: added support for copying audio and subtitle streams
* fix: fixed syntax issues for MSVC
* fix: fixed audio/subtitle timestamp rescaling

Signed-off-by: k4yt3x <i@k4yt3x.com>

include/conversions.h

@@ -1,7 +1,10 @@
 #ifndef CONVERSIONS_H
 #define CONVERSIONS_H
 
-#include <libavutil/frame.h>
+extern "C" {
+#include <libswscale/swscale.h>
+}
+
 #include <mat.h>
 
 // Convert AVFrame to another pixel format

include/decoder.h

@@ -1,11 +1,14 @@
 #ifndef DECODER_H
 #define DECODER_H
 
+extern "C" {
 #include <libavcodec/avcodec.h>
-#include <libavfilter/avfilter.h>
 #include <libavformat/avformat.h>
+}
 
 int init_decoder(
+    AVHWDeviceType hw_type,
+    AVBufferRef *hw_ctx,
     const char *input_filename,
     AVFormatContext **fmt_ctx,
     AVCodecContext **dec_ctx,

include/encoder.h

@@ -1,20 +1,32 @@
 #ifndef ENCODER_H
 #define ENCODER_H
 
+extern "C" {
 #include <libavcodec/avcodec.h>
 #include <libavformat/avformat.h>
+#include <libavutil/opt.h>
+}
 
 #include "libvideo2x.h"
 
 int init_encoder(
+    AVBufferRef *hw_ctx,
     const char *output_filename,
+    AVFormatContext *ifmt_ctx,
     AVFormatContext **ofmt_ctx,
     AVCodecContext **enc_ctx,
     AVCodecContext *dec_ctx,
-    EncoderConfig *encoder_config
+    EncoderConfig *encoder_config,
+    int video_stream_index,
+    int **stream_mapping
 );
 
-int encode_and_write_frame(AVFrame *frame, AVCodecContext *enc_ctx, AVFormatContext *ofmt_ctx);
+int encode_and_write_frame(
+    AVFrame *frame,
+    AVCodecContext *enc_ctx,
+    AVFormatContext *ofmt_ctx,
+    int video_stream_index
+);
 
 int flush_encoder(AVCodecContext *enc_ctx, AVFormatContext *ofmt_ctx);
 

include/filter.h

@@ -6,14 +6,15 @@
 extern "C" {
 #include <libavcodec/avcodec.h>
 #include <libavfilter/avfilter.h>
+#include <libavutil/buffer.h>
 }
 
 // Abstract base class for filters
 class Filter {
    public:
     virtual ~Filter() {}
-    virtual int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx) = 0;
-    virtual AVFrame *process_frame(AVFrame *input_frame) = 0;
+    virtual int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx, AVBufferRef *hw_ctx) = 0;
+    virtual int process_frame(AVFrame *input_frame, AVFrame **output_frame) = 0;
     virtual int flush(std::vector<AVFrame *> &processed_frames) = 0;
 };
 

include/getopt.h

@@ -22,7 +22,7 @@ struct option {
 #define required_argument 1
 #define optional_argument 2
 
-int getopt(int, char **, const char *);
+// int getopt(int, char **, const char *);
 int getopt_long(int, char **, const char *, const struct option *, int *);
 
 #ifdef __cplusplus

include/libplacebo.h

@@ -3,15 +3,16 @@
 
 #include <filesystem>
 
+extern "C" {
 #include <libavcodec/avcodec.h>
 #include <libavfilter/avfilter.h>
-#include <libavutil/buffer.h>
+}
 
 int init_libplacebo(
+    AVBufferRef *hw_ctx,
     AVFilterGraph **filter_graph,
     AVFilterContext **buffersrc_ctx,
     AVFilterContext **buffersink_ctx,
-    AVBufferRef **device_ctx,
     AVCodecContext *dec_ctx,
     int output_width,
     int output_height,

include/libplacebo_filter.h

@@ -3,7 +3,11 @@
 
 #include <filesystem>
 
-#include <libavutil/buffer.h>
+extern "C" {
+#include <libavcodec/avcodec.h>
+#include <libavfilter/buffersink.h>
+#include <libavfilter/buffersrc.h>
+}
 
 #include "filter.h"
 
@@ -13,7 +17,6 @@ class LibplaceboFilter : public Filter {
     AVFilterGraph *filter_graph;
     AVFilterContext *buffersrc_ctx;
     AVFilterContext *buffersink_ctx;
-    AVBufferRef *device_ctx;
     int output_width;
     int output_height;
     const std::filesystem::path shader_path;
@@ -27,10 +30,10 @@ class LibplaceboFilter : public Filter {
     virtual ~LibplaceboFilter();
 
     // Initializes the filter with decoder and encoder contexts
-    int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx) override;
+    int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx, AVBufferRef *hw_ctx) override;
 
     // Processes an input frame and returns the processed frame
-    AVFrame *process_frame(AVFrame *input_frame) override;
+    int process_frame(AVFrame *input_frame, AVFrame **output_frame) override;
 
     // Flushes any remaining frames
     int flush(std::vector<AVFrame *> &processed_frames) override;

include/libvideo2x.h

@@ -1,13 +1,10 @@
 #ifndef LIBVIDEO2X_H
 #define LIBVIDEO2X_H
 
-#include <libavutil/pixfmt.h>
+#include <stdbool.h>
 #include <stdint.h>
 #include <time.h>
 
-#include <libavcodec/avcodec.h>
-#include <libavcodec/codec_id.h>
-
 #ifdef _WIN32
 #ifdef LIBVIDEO2X_EXPORTS
 #define LIBVIDEO2X_API __declspec(dllexport)
@@ -22,6 +19,9 @@
 extern "C" {
 #endif
 
+#include <libavcodec/avcodec.h>
+#include <libavformat/avformat.h>
+
 // Enum to specify filter type
 enum FilterType {
     FILTER_LIBPLACEBO,
@@ -38,7 +38,7 @@ struct LibplaceboConfig {
 // Configuration for RealESRGAN filter
 struct RealESRGANConfig {
     int gpuid;
-    int tta_mode;
+    bool tta_mode;
     int scaling_factor;
     const char *model;
 };
@@ -56,6 +56,7 @@ struct FilterConfig {
 struct EncoderConfig {
     int output_width;
     int output_height;
+    bool copy_streams;
     enum AVCodecID codec;
     enum AVPixelFormat pix_fmt;
     const char *preset;
@@ -63,20 +64,25 @@ struct EncoderConfig {
     float crf;
 };
 
-// Processing status
-struct ProcessingStatus {
+// Video processing context
+struct VideoProcessingContext {
     int64_t processed_frames;
     int64_t total_frames;
     time_t start_time;
+    bool pause;
+    bool abort;
+    bool completed;
 };
 
 // C-compatible process_video function
 LIBVIDEO2X_API int process_video(
     const char *input_filename,
     const char *output_filename,
+    bool benchmark,
+    enum AVHWDeviceType hw_device_type,
     const struct FilterConfig *filter_config,
     struct EncoderConfig *encoder_config,
-    struct ProcessingStatus *status
+    struct VideoProcessingContext *proc_ctx
 );
 
 #ifdef __cplusplus

include/realesrgan_filter.h

@@ -3,6 +3,10 @@
 
 #include <filesystem>
 
+extern "C" {
+#include <libavcodec/avcodec.h>
+}
+
 #include "filter.h"
 #include "realesrgan.h"
 
@@ -35,10 +39,10 @@ class RealesrganFilter : public Filter {
     virtual ~RealesrganFilter();
 
     // Initializes the filter with decoder and encoder contexts
-    int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx) override;
+    int init(AVCodecContext *dec_ctx, AVCodecContext *enc_ctx, AVBufferRef *hw_ctx) override;
 
     // Processes an input frame and returns the processed frame
-    AVFrame *process_frame(AVFrame *input_frame) override;
+    int process_frame(AVFrame *input_frame, AVFrame **output_frame) override;
 
     // Flushes any remaining frames (if necessary)
     int flush(std::vector<AVFrame *> &processed_frames) override;