1233d2500723e5594f3e7c70896ffeeef32b9c950ywan/*
2233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3233d2500723e5594f3e7c70896ffeeef32b9c950ywan *
4233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  Use of this source code is governed by a BSD-style license
5233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  that can be found in the LICENSE file in the root of the source
6233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  tree. An additional intellectual property rights grant can be found
7233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  in the file PATENTS.  All contributing project authors may
8233d2500723e5594f3e7c70896ffeeef32b9c950ywan *  be found in the AUTHORS file in the root of the source tree.
9233d2500723e5594f3e7c70896ffeeef32b9c950ywan */
10233d2500723e5594f3e7c70896ffeeef32b9c950ywan
11233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Two Pass Encoder
12233d2500723e5594f3e7c70896ffeeef32b9c950ywan// ================
13233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
14233d2500723e5594f3e7c70896ffeeef32b9c950ywan// This is an example of a two pass encoder loop. It takes an input file in
15233d2500723e5594f3e7c70896ffeeef32b9c950ywan// YV12 format, passes it through the encoder twice, and writes the compressed
16233d2500723e5594f3e7c70896ffeeef32b9c950ywan// frames to disk in IVF format. It builds upon the simple_encoder example.
17233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
18233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Twopass Variables
19233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -----------------
20233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Twopass mode needs to track the current pass number and the buffer of
21233d2500723e5594f3e7c70896ffeeef32b9c950ywan// statistics packets.
22233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
23233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Updating The Configuration
24233d2500723e5594f3e7c70896ffeeef32b9c950ywan// ---------------------------------
25233d2500723e5594f3e7c70896ffeeef32b9c950ywan// In two pass mode, the configuration has to be updated on each pass. The
26233d2500723e5594f3e7c70896ffeeef32b9c950ywan// statistics buffer is passed on the last pass.
27233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
28233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Encoding A Frame
29233d2500723e5594f3e7c70896ffeeef32b9c950ywan// ----------------
30233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Encoding a frame in two pass mode is identical to the simple encoder
31233d2500723e5594f3e7c70896ffeeef32b9c950ywan// example, except the deadline is set to VPX_DL_BEST_QUALITY to get the
32233d2500723e5594f3e7c70896ffeeef32b9c950ywan// best quality possible. VPX_DL_GOOD_QUALITY could also be used.
33233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
34233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
35233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Processing Statistics Packets
36233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -----------------------------
37233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Each packet of type `VPX_CODEC_CX_FRAME_PKT` contains the encoded data
38233d2500723e5594f3e7c70896ffeeef32b9c950ywan// for this frame. We write a IVF frame header, followed by the raw data.
39233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
40233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
41233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Pass Progress Reporting
42233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -----------------------------
43233d2500723e5594f3e7c70896ffeeef32b9c950ywan// It's sometimes helpful to see when each pass completes.
44233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
45233d2500723e5594f3e7c70896ffeeef32b9c950ywan//
46233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Clean-up
47233d2500723e5594f3e7c70896ffeeef32b9c950ywan// -----------------------------
48233d2500723e5594f3e7c70896ffeeef32b9c950ywan// Destruction of the encoder instance must be done on each pass. The
49233d2500723e5594f3e7c70896ffeeef32b9c950ywan// raw image should be destroyed at the end as usual.
50233d2500723e5594f3e7c70896ffeeef32b9c950ywan
51233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <stdio.h>
52233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <stdlib.h>
53233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include <string.h>
54233d2500723e5594f3e7c70896ffeeef32b9c950ywan
55233d2500723e5594f3e7c70896ffeeef32b9c950ywan#define VPX_CODEC_DISABLE_COMPAT 1
56233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "vpx/vpx_encoder.h"
57233d2500723e5594f3e7c70896ffeeef32b9c950ywan
58233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "./tools_common.h"
59233d2500723e5594f3e7c70896ffeeef32b9c950ywan#include "./video_writer.h"
60233d2500723e5594f3e7c70896ffeeef32b9c950ywan
61233d2500723e5594f3e7c70896ffeeef32b9c950ywanstatic const char *exec_name;
62233d2500723e5594f3e7c70896ffeeef32b9c950ywan
63233d2500723e5594f3e7c70896ffeeef32b9c950ywanvoid usage_exit() {
64233d2500723e5594f3e7c70896ffeeef32b9c950ywan  fprintf(stderr, "Usage: %s <codec> <width> <height> <infile> <outfile>\n",
65233d2500723e5594f3e7c70896ffeeef32b9c950ywan          exec_name);
66233d2500723e5594f3e7c70896ffeeef32b9c950ywan  exit(EXIT_FAILURE);
67233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
68233d2500723e5594f3e7c70896ffeeef32b9c950ywan
69233d2500723e5594f3e7c70896ffeeef32b9c950ywanstatic void get_frame_stats(vpx_codec_ctx_t *ctx,
70233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            const vpx_image_t *img,
71233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            vpx_codec_pts_t pts,
72233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            unsigned int duration,
73233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            vpx_enc_frame_flags_t flags,
74233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            unsigned int deadline,
75233d2500723e5594f3e7c70896ffeeef32b9c950ywan                            vpx_fixed_buf_t *stats) {
76233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_iter_t iter = NULL;
77233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_cx_pkt_t *pkt = NULL;
78233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_err_t res = vpx_codec_encode(ctx, img, pts, duration, flags,
79233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                               deadline);
80233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (res != VPX_CODEC_OK)
81233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(ctx, "Failed to get frame stats.");
82233d2500723e5594f3e7c70896ffeeef32b9c950ywan
83233d2500723e5594f3e7c70896ffeeef32b9c950ywan  while ((pkt = vpx_codec_get_cx_data(ctx, &iter)) != NULL) {
84233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (pkt->kind == VPX_CODEC_STATS_PKT) {
85233d2500723e5594f3e7c70896ffeeef32b9c950ywan      const uint8_t *const pkt_buf = pkt->data.twopass_stats.buf;
86233d2500723e5594f3e7c70896ffeeef32b9c950ywan      const size_t pkt_size = pkt->data.twopass_stats.sz;
87233d2500723e5594f3e7c70896ffeeef32b9c950ywan      stats->buf = realloc(stats->buf, stats->sz + pkt_size);
88233d2500723e5594f3e7c70896ffeeef32b9c950ywan      memcpy((uint8_t *)stats->buf + stats->sz, pkt_buf, pkt_size);
89233d2500723e5594f3e7c70896ffeeef32b9c950ywan      stats->sz += pkt_size;
90233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
91233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
92233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
93233d2500723e5594f3e7c70896ffeeef32b9c950ywan
94233d2500723e5594f3e7c70896ffeeef32b9c950ywanstatic void encode_frame(vpx_codec_ctx_t *ctx,
95233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         const vpx_image_t *img,
96233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         vpx_codec_pts_t pts,
97233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         unsigned int duration,
98233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         vpx_enc_frame_flags_t flags,
99233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         unsigned int deadline,
100233d2500723e5594f3e7c70896ffeeef32b9c950ywan                         VpxVideoWriter *writer) {
101233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_iter_t iter = NULL;
102233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_cx_pkt_t *pkt = NULL;
103233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const vpx_codec_err_t res = vpx_codec_encode(ctx, img, pts, duration, flags,
104233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                               deadline);
105233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (res != VPX_CODEC_OK)
106233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(ctx, "Failed to encode frame.");
107233d2500723e5594f3e7c70896ffeeef32b9c950ywan
108233d2500723e5594f3e7c70896ffeeef32b9c950ywan  while ((pkt = vpx_codec_get_cx_data(ctx, &iter)) != NULL) {
109233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (pkt->kind == VPX_CODEC_CX_FRAME_PKT) {
110233d2500723e5594f3e7c70896ffeeef32b9c950ywan      const int keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0;
111233d2500723e5594f3e7c70896ffeeef32b9c950ywan
112233d2500723e5594f3e7c70896ffeeef32b9c950ywan      if (!vpx_video_writer_write_frame(writer, pkt->data.frame.buf,
113233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                                pkt->data.frame.sz,
114233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                                pkt->data.frame.pts))
115233d2500723e5594f3e7c70896ffeeef32b9c950ywan        die_codec(ctx, "Failed to write compressed frame.");
116233d2500723e5594f3e7c70896ffeeef32b9c950ywan      printf(keyframe ? "K" : ".");
117233d2500723e5594f3e7c70896ffeeef32b9c950ywan      fflush(stdout);
118233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
119233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
120233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
121233d2500723e5594f3e7c70896ffeeef32b9c950ywan
122233d2500723e5594f3e7c70896ffeeef32b9c950ywanint main(int argc, char **argv) {
123233d2500723e5594f3e7c70896ffeeef32b9c950ywan  FILE *infile = NULL;
124233d2500723e5594f3e7c70896ffeeef32b9c950ywan  VpxVideoWriter *writer = NULL;
125233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_ctx_t codec;
126233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_enc_cfg_t cfg;
127233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_image_t raw;
128233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_codec_err_t res;
129233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_fixed_buf_t stats = {0};
130233d2500723e5594f3e7c70896ffeeef32b9c950ywan  VpxVideoInfo info = {0};
131233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const VpxInterface *encoder = NULL;
132233d2500723e5594f3e7c70896ffeeef32b9c950ywan  int pass;
133233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const int fps = 30;        // TODO(dkovalev) add command line argument
134233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const int bitrate = 200;   // kbit/s TODO(dkovalev) add command line argument
135233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const char *const codec_arg = argv[1];
136233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const char *const width_arg = argv[2];
137233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const char *const height_arg = argv[3];
138233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const char *const infile_arg = argv[4];
139233d2500723e5594f3e7c70896ffeeef32b9c950ywan  const char *const outfile_arg = argv[5];
140233d2500723e5594f3e7c70896ffeeef32b9c950ywan  exec_name = argv[0];
141233d2500723e5594f3e7c70896ffeeef32b9c950ywan
142233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (argc != 6)
143233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Invalid number of arguments.");
144233d2500723e5594f3e7c70896ffeeef32b9c950ywan
145233d2500723e5594f3e7c70896ffeeef32b9c950ywan  encoder = get_vpx_encoder_by_name(codec_arg);
146233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!encoder)
147233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Unsupported codec.");
148233d2500723e5594f3e7c70896ffeeef32b9c950ywan
149233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.codec_fourcc = encoder->fourcc;
150233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.time_base.numerator = 1;
151233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.time_base.denominator = fps;
152233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.frame_width = strtol(width_arg, NULL, 0);
153233d2500723e5594f3e7c70896ffeeef32b9c950ywan  info.frame_height = strtol(height_arg, NULL, 0);
154233d2500723e5594f3e7c70896ffeeef32b9c950ywan
155233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (info.frame_width <= 0 ||
156233d2500723e5594f3e7c70896ffeeef32b9c950ywan      info.frame_height <= 0 ||
157233d2500723e5594f3e7c70896ffeeef32b9c950ywan      (info.frame_width % 2) != 0 ||
158233d2500723e5594f3e7c70896ffeeef32b9c950ywan      (info.frame_height % 2) != 0) {
159233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Invalid frame size: %dx%d", info.frame_width, info.frame_height);
160233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
161233d2500723e5594f3e7c70896ffeeef32b9c950ywan
162233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, info.frame_width,
163233d2500723e5594f3e7c70896ffeeef32b9c950ywan                                             info.frame_height, 1)) {
164233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Failed to allocate image", info.frame_width, info.frame_height);
165233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
166233d2500723e5594f3e7c70896ffeeef32b9c950ywan
167233d2500723e5594f3e7c70896ffeeef32b9c950ywan  writer = vpx_video_writer_open(outfile_arg, kContainerIVF, &info);
168233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (!writer)
169233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die("Failed to open %s for writing", outfile_arg);
170233d2500723e5594f3e7c70896ffeeef32b9c950ywan
171233d2500723e5594f3e7c70896ffeeef32b9c950ywan  printf("Using %s\n", vpx_codec_iface_name(encoder->interface()));
172233d2500723e5594f3e7c70896ffeeef32b9c950ywan
173233d2500723e5594f3e7c70896ffeeef32b9c950ywan  res = vpx_codec_enc_config_default(encoder->interface(), &cfg, 0);
174233d2500723e5594f3e7c70896ffeeef32b9c950ywan  if (res)
175233d2500723e5594f3e7c70896ffeeef32b9c950ywan    die_codec(&codec, "Failed to get default codec config.");
176233d2500723e5594f3e7c70896ffeeef32b9c950ywan
177233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_w = info.frame_width;
178233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_h = info.frame_height;
179233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_timebase.num = info.time_base.numerator;
180233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.g_timebase.den = info.time_base.denominator;
181233d2500723e5594f3e7c70896ffeeef32b9c950ywan  cfg.rc_target_bitrate = bitrate;
182233d2500723e5594f3e7c70896ffeeef32b9c950ywan
183233d2500723e5594f3e7c70896ffeeef32b9c950ywan  for (pass = 0; pass < 2; ++pass) {
184233d2500723e5594f3e7c70896ffeeef32b9c950ywan    int frame_count = 0;
185233d2500723e5594f3e7c70896ffeeef32b9c950ywan
186233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (pass == 0) {
187233d2500723e5594f3e7c70896ffeeef32b9c950ywan      cfg.g_pass = VPX_RC_FIRST_PASS;
188233d2500723e5594f3e7c70896ffeeef32b9c950ywan    } else {
189233d2500723e5594f3e7c70896ffeeef32b9c950ywan      cfg.g_pass = VPX_RC_LAST_PASS;
190233d2500723e5594f3e7c70896ffeeef32b9c950ywan      cfg.rc_twopass_stats_in = stats;
191233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
192233d2500723e5594f3e7c70896ffeeef32b9c950ywan
193233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (!(infile = fopen(infile_arg, "rb")))
194233d2500723e5594f3e7c70896ffeeef32b9c950ywan      die("Failed to open %s for reading", infile_arg);
195233d2500723e5594f3e7c70896ffeeef32b9c950ywan
196233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (vpx_codec_enc_init(&codec, encoder->interface(), &cfg, 0))
197233d2500723e5594f3e7c70896ffeeef32b9c950ywan      die_codec(&codec, "Failed to initialize encoder");
198233d2500723e5594f3e7c70896ffeeef32b9c950ywan
199233d2500723e5594f3e7c70896ffeeef32b9c950ywan    while (vpx_img_read(&raw, infile)) {
200233d2500723e5594f3e7c70896ffeeef32b9c950ywan      ++frame_count;
201233d2500723e5594f3e7c70896ffeeef32b9c950ywan
202233d2500723e5594f3e7c70896ffeeef32b9c950ywan      if (pass == 0) {
203233d2500723e5594f3e7c70896ffeeef32b9c950ywan        get_frame_stats(&codec, &raw, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
204233d2500723e5594f3e7c70896ffeeef32b9c950ywan                        &stats);
205233d2500723e5594f3e7c70896ffeeef32b9c950ywan      } else {
206233d2500723e5594f3e7c70896ffeeef32b9c950ywan        encode_frame(&codec, &raw, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
207233d2500723e5594f3e7c70896ffeeef32b9c950ywan                     writer);
208233d2500723e5594f3e7c70896ffeeef32b9c950ywan      }
209233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
210233d2500723e5594f3e7c70896ffeeef32b9c950ywan
211233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (pass == 0) {
212233d2500723e5594f3e7c70896ffeeef32b9c950ywan      get_frame_stats(&codec, NULL, frame_count, 1, 0, VPX_DL_BEST_QUALITY,
213233d2500723e5594f3e7c70896ffeeef32b9c950ywan                      &stats);
214233d2500723e5594f3e7c70896ffeeef32b9c950ywan    } else {
215233d2500723e5594f3e7c70896ffeeef32b9c950ywan      printf("\n");
216233d2500723e5594f3e7c70896ffeeef32b9c950ywan    }
217233d2500723e5594f3e7c70896ffeeef32b9c950ywan
218233d2500723e5594f3e7c70896ffeeef32b9c950ywan    fclose(infile);
219233d2500723e5594f3e7c70896ffeeef32b9c950ywan    printf("Pass %d complete. Processed %d frames.\n", pass + 1, frame_count);
220233d2500723e5594f3e7c70896ffeeef32b9c950ywan    if (vpx_codec_destroy(&codec))
221233d2500723e5594f3e7c70896ffeeef32b9c950ywan      die_codec(&codec, "Failed to destroy codec.");
222233d2500723e5594f3e7c70896ffeeef32b9c950ywan  }
223233d2500723e5594f3e7c70896ffeeef32b9c950ywan
224233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_img_free(&raw);
225233d2500723e5594f3e7c70896ffeeef32b9c950ywan  free(stats.buf);
226233d2500723e5594f3e7c70896ffeeef32b9c950ywan
227233d2500723e5594f3e7c70896ffeeef32b9c950ywan  vpx_video_writer_close(writer);
228233d2500723e5594f3e7c70896ffeeef32b9c950ywan
229233d2500723e5594f3e7c70896ffeeef32b9c950ywan  return EXIT_SUCCESS;
230233d2500723e5594f3e7c70896ffeeef32b9c950ywan}
231