libavcodec transcoding - setting pts and dts - c

I'm trying to transcode video files into a standard format. This is the main loop I have:
for(frame_count = 0; av_read_frame(input_ctx, &in_packet) >= 0; frame_count++) {
if(in_packet.stream_index == video_stream_index) {
decodedPacket = 0;
rc = avcodec_decode_video2(video_in_codec, inputFrame, &decodedPacket, &in_packet);
if(decodedPacket) {
out_frames++;
rc = sws_scale(sws_ctx, inputFrame->data, inputFrame->linesize, 0, video_out_codec->height, outputFrame->data, outputFrame->linesize);
if(rc != video_out_codec->height) {
puts("scaling error");
}
outputFrame->pts = (1.0 / 30.0) * 90.0 * video_out_codec->frame_number;
rc = avcodec_encode_video2(video_out_codec, &out_packet, outputFrame, &encodedPacket);
if(rc != 0) {
puts("encoding error");
}
if(encodedPacket) {
if (video_out_stream->codec->coded_frame->key_frame) { // deprecated, what to use instead
out_packet.flags |= AV_PKT_FLAG_KEY;
}
out_packet.stream_index = 0;
rc = av_interleaved_write_frame(output_ctx, &out_packet);
if(rc != 0) {
puts("frame write error");
}
av_free_packet(&out_packet);
memset(&out_packet, 0, sizeof(AVPacket));
packet_count++;
}
av_free_packet(&in_packet);
}
}
...
When I run this I get strange values for tbr/tbn/tbc reported by ffprobe and the video is a mess.
I've tried adding code as per this answer e.g.
if (out_packet.pts != AV_NOPTS_VALUE) {
out_packet.pts = av_rescale_q(out_packet.pts, video_out_stream->codec->time_base, video_out_stream->time_base);
}
if (out_packet.pts != AV_NOPTS_VALUE) {
out_packet.dts = av_rescale_q(out_packet.dts, video_out_stream->codec->time_base, video_out_stream->time_base);
}
...but then I get errors such as this is the debug output:
[mp4 # 0x1038aba00] Delay between the first packet and last packet in the muxing queue is 10100000 > 10000000: forcing output
I'm setting the timebase correctly (I think)...
video_out_stream = avformat_new_stream(output_ctx, videoEncoder);
video_out_stream->id = 0;
video_out_stream->time_base.den = 30;
video_out_stream->time_base.num = 1;
video_out_codec = video_out_stream->codec;
avcodec_get_context_defaults3(video_out_codec, videoEncoder);
video_out_codec->codec_id = AV_CODEC_ID_H264;
video_out_codec->bit_rate = 2048;
video_out_codec->width = 854;
video_out_codec->height = 480;
video_out_codec->time_base.den = 30;
video_out_codec->time_base.num = 1;
video_out_codec->gop_size = 30;
video_out_codec->pix_fmt = AV_PIX_FMT_YUV420P;
Any thoughts about how I can calculate the correct output frame pts and encoded packet dts/pts? I'm clearly doing something wrong.

Related

Vpx temporal svc decoder

I would like to test libvpx (vp9) temporal svc. For that matter I take examples/vpx_temporal_svc_encoder.c and modify it to add a decoder. If I take for example mode 3 with 3 temporal layers and 6 frame period and try to decode only layer 0 or layers 0 and 1 I get an error: "Failed to decode frame.: Corrupt frame detected. Failed to decode tile data". What am I missing ? Here is my code (had to remove unmodified part due to post limits):
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
// This is an example demonstrating how to implement a multi-layer VPx
// encoding scheme based on temporal scalability for video applications
// that benefit from a scalable bitstream.
#include <assert.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "./vpx_config.h"
#include "./y4minput.h"
#include "../vpx_ports/vpx_timer.h"
#include "vpx/vp8cx.h"
#include "vpx/vpx_encoder.h"
#include "vpx_ports/bitops.h"
#include "../tools_common.h"
#include "../video_writer.h"
#include "vpx/vpx_decoder.h"
#define ROI_MAP 0
#define zero(Dest) memset(&(Dest), 0, sizeof(Dest));
static const char *exec_name;
void usage_exit(void) { exit(EXIT_FAILURE); }
// Denoiser states for vp8, for temporal denoising.
enum denoiserStateVp8 {
kVp8DenoiserOff,
kVp8DenoiserOnYOnly,
kVp8DenoiserOnYUV,
kVp8DenoiserOnYUVAggressive,
kVp8DenoiserOnAdaptive
};
// Denoiser states for vp9, for temporal denoising.
enum denoiserStateVp9 {
kVp9DenoiserOff,
kVp9DenoiserOnYOnly,
// For SVC: denoise the top two spatial layers.
kVp9DenoiserOnYTwoSpatialLayers
};
static int mode_to_num_layers[13] = { 1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3, 3 };
int main(int argc, char **argv) {
VpxVideoWriter *outfile[VPX_TS_MAX_LAYERS] = { NULL };
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t cfg;
int frame_cnt = 0;
vpx_image_t raw;
vpx_codec_err_t res;
unsigned int width;
unsigned int height;
uint32_t error_resilient = 0;
int speed;
int frame_avail;
int got_data;
int flags = 0;
unsigned int i;
int pts = 0; // PTS starts at 0.
int frame_duration = 1; // 1 timebase tick per frame.
int layering_mode = 0;
int layer_flags[VPX_TS_MAX_PERIODICITY] = { 0 };
int flag_periodicity = 1;
#if ROI_MAP
vpx_roi_map_t roi;
#endif
vpx_svc_layer_id_t layer_id;
const VpxInterface *encoder = NULL;
struct VpxInputContext input_ctx;
struct RateControlMetrics rc;
int64_t cx_time = 0;
const int min_args_base = 13;
#if CONFIG_VP9_HIGHBITDEPTH
vpx_bit_depth_t bit_depth = VPX_BITS_8;
int input_bit_depth = 8;
const int min_args = min_args_base + 1;
#else
const int min_args = min_args_base;
#endif // CONFIG_VP9_HIGHBITDEPTH
double sum_bitrate = 0.0;
double sum_bitrate2 = 0.0;
double framerate = 30.0;
zero(rc.layer_target_bitrate);
memset(&layer_id, 0, sizeof(vpx_svc_layer_id_t));
memset(&input_ctx, 0, sizeof(input_ctx));
/* Setup default input stream settings */
input_ctx.framerate.numerator = 30;
input_ctx.framerate.denominator = 1;
input_ctx.only_i420 = 1;
input_ctx.bit_depth = 0;
exec_name = argv[0];
// Check usage and arguments.
if (argc < min_args) {
#if CONFIG_VP9_HIGHBITDEPTH
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> "
"<error_resilient> <threads> <mode> "
"<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n",
argv[0]);
#else
//./vpx_temporal_svc_encoder ../9.yuv svc vp9 256 256 1 25 9 0 0 1 3 500000 500000 50000
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> "
"<error_resilient> <threads> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n",
argv[0]);
#endif // CONFIG_VP9_HIGHBITDEPTH
}
encoder = get_vpx_encoder_by_name(argv[3]);
if (!encoder) die("Unsupported codec.");
printf("Using %s\n", vpx_codec_iface_name(encoder->codec_interface()));
width = (unsigned int)strtoul(argv[4], NULL, 0);
height = (unsigned int)strtoul(argv[5], NULL, 0);
if (width < 16 || width % 2 || height < 16 || height % 2) {
die("Invalid resolution: %d x %d", width, height);
}
layering_mode = (int)strtol(argv[12], NULL, 0);
if (layering_mode < 0 || layering_mode > 13) {
die("Invalid layering mode (0..12) %s", argv[12]);
}
if (argc != min_args + mode_to_num_layers[layering_mode]) {
die("Invalid number of arguments");
}
input_ctx.filename = argv[1];
open_input_file(&input_ctx);
#if CONFIG_VP9_HIGHBITDEPTH
switch (strtol(argv[argc - 1], NULL, 0)) {
case 8:
bit_depth = VPX_BITS_8;
input_bit_depth = 8;
break;
case 10:
bit_depth = VPX_BITS_10;
input_bit_depth = 10;
break;
case 12:
bit_depth = VPX_BITS_12;
input_bit_depth = 12;
break;
default: die("Invalid bit depth (8, 10, 12) %s", argv[argc - 1]);
}
// Y4M reader has its own allocation.
if (input_ctx.file_type != FILE_TYPE_Y4M) {
if (!vpx_img_alloc(
&raw,
bit_depth == VPX_BITS_8 ? VPX_IMG_FMT_I420 : VPX_IMG_FMT_I42016,
width, height, 32)) {
die("Failed to allocate image", width, height);
}
}
#else
// Y4M reader has its own allocation.
if (input_ctx.file_type != FILE_TYPE_Y4M) {
if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, width, height, 32)) {
die("Failed to allocate image", width, height);
}
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Populate encoder configuration.
res = vpx_codec_enc_config_default(encoder->codec_interface(), &cfg, 0);
if (res) {
printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
return EXIT_FAILURE;
}
// Update the default configuration with our settings.
cfg.g_w = width;
cfg.g_h = height;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth != VPX_BITS_8) {
cfg.g_bit_depth = bit_depth;
cfg.g_input_bit_depth = input_bit_depth;
cfg.g_profile = 2;
}
#endif // CONFIG_VP9_HIGHBITDEPTH
// Timebase format e.g. 30fps: numerator=1, demoninator = 30.
cfg.g_timebase.num = (int)strtol(argv[6], NULL, 0);
cfg.g_timebase.den = (int)strtol(argv[7], NULL, 0);
speed = (int)strtol(argv[8], NULL, 0);
if (speed < 0) {
die("Invalid speed setting: must be positive");
}
if (strncmp(encoder->name, "vp9", 3) == 0 && speed > 9) {
warn("Mapping speed %d to speed 9.\n", speed);
}
for (i = min_args_base;
(int)i < min_args_base + mode_to_num_layers[layering_mode]; ++i) {
rc.layer_target_bitrate[i - 13] = (int)strtol(argv[i], NULL, 0);
if (strncmp(encoder->name, "vp8", 3) == 0)
cfg.ts_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
else if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.layer_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
}
// Real time parameters.
cfg.rc_dropframe_thresh = (unsigned int)strtoul(argv[9], NULL, 0);
cfg.rc_end_usage = VPX_CBR;
cfg.rc_min_quantizer = 2;
cfg.rc_max_quantizer = 56;
if (strncmp(encoder->name, "vp9", 3) == 0) cfg.rc_max_quantizer = 52;
cfg.rc_undershoot_pct = 50;
cfg.rc_overshoot_pct = 50;
cfg.rc_buf_initial_sz = 600;
cfg.rc_buf_optimal_sz = 600;
cfg.rc_buf_sz = 1000;
// Disable dynamic resizing by default.
cfg.rc_resize_allowed = 0;
// Use 1 thread as default.
cfg.g_threads = (unsigned int)strtoul(argv[11], NULL, 0);
error_resilient = (uint32_t)strtoul(argv[10], NULL, 0);
if (error_resilient != 0 && error_resilient != 1) {
die("Invalid value for error resilient (0, 1): %d.", error_resilient);
}
// Enable error resilient mode.
cfg.g_error_resilient = error_resilient;
cfg.g_lag_in_frames = 0;
cfg.kf_mode = VPX_KF_AUTO;
// Disable automatic keyframe placement.
cfg.kf_min_dist = cfg.kf_max_dist = 3000;
cfg.temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
set_temporal_layer_pattern(layering_mode, &cfg, layer_flags,
&flag_periodicity);
set_rate_control_metrics(&rc, &cfg);
if (input_ctx.file_type == FILE_TYPE_Y4M) {
if (input_ctx.width != cfg.g_w || input_ctx.height != cfg.g_h) {
die("Incorrect width or height: %d x %d", cfg.g_w, cfg.g_h);
}
if (input_ctx.framerate.numerator != cfg.g_timebase.den ||
input_ctx.framerate.denominator != cfg.g_timebase.num) {
die("Incorrect framerate: numerator %d denominator %d",
cfg.g_timebase.num, cfg.g_timebase.den);
}
}
framerate = cfg.g_timebase.den / cfg.g_timebase.num;
// Open an output file for each stream.
for (i = 0; i < cfg.ts_number_layers; ++i) {
char file_name[PATH_MAX];
VpxVideoInfo info;
info.codec_fourcc = encoder->fourcc;
info.frame_width = cfg.g_w;
info.frame_height = cfg.g_h;
info.time_base.numerator = cfg.g_timebase.num;
info.time_base.denominator = cfg.g_timebase.den;
snprintf(file_name, sizeof(file_name), "%s_%d.ivf", argv[2], i);
outfile[i] = vpx_video_writer_open(file_name, kContainerIVF, &info);
if (!outfile[i]) die("Failed to open %s for writing", file_name);
assert(outfile[i] != NULL);
}
// No spatial layers in this encoder.
cfg.ss_number_layers = 1;
// Initialize codec.
#if CONFIG_VP9_HIGHBITDEPTH
if (vpx_codec_enc_init(
&codec, encoder->codec_interface(), &cfg,
bit_depth == VPX_BITS_8 ? 0 : VPX_CODEC_USE_HIGHBITDEPTH))
#else
if (vpx_codec_enc_init(&codec, encoder->codec_interface(), &cfg, 0))
#endif // CONFIG_VP9_HIGHBITDEPTH
die("Failed to initialize encoder");
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kVp8DenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP8E_SET_GF_CBR_BOOST_PCT, 0);
#if ROI_MAP
set_roi_map(encoder->name, &cfg, &roi);
if (vpx_codec_control(&codec, VP8E_SET_ROI_MAP, &roi))
die_codec(&codec, "Failed to set ROI map");
#endif
} else if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_svc_extra_cfg_t svc_params;
memset(&svc_params, 0, sizeof(svc_params));
vpx_codec_control(&codec, VP9E_SET_POSTENCODE_DROP, 0);
vpx_codec_control(&codec, VP9E_SET_DISABLE_OVERSHOOT_MAXQ_CBR, 0);
vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
vpx_codec_control(&codec, VP9E_SET_GF_CBR_BOOST_PCT, 0);
vpx_codec_control(&codec, VP9E_SET_FRAME_PARALLEL_DECODING, 0);
vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kVp9DenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP9E_SET_TUNE_CONTENT, 0);
vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, get_msb(cfg.g_threads));
vpx_codec_control(&codec, VP9E_SET_DISABLE_LOOPFILTER, 0);
#if ROI_MAP
set_roi_map(encoder->name, &cfg, &roi);
if (vpx_codec_control(&codec, VP9E_SET_ROI_MAP, &roi))
die_codec(&codec, "Failed to set ROI map");
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 0);
#endif
if (cfg.g_threads > 1)
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 1);
else
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 0);
if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1 : 0))
die_codec(&codec, "Failed to set SVC");
for (i = 0; i < cfg.ts_number_layers; ++i) {
svc_params.max_quantizers[i] = cfg.rc_max_quantizer;
svc_params.min_quantizers[i] = cfg.rc_min_quantizer;
}
svc_params.scaling_factor_num[0] = cfg.g_h;
svc_params.scaling_factor_den[0] = cfg.g_h;
vpx_codec_control(&codec, VP9E_SET_SVC_PARAMETERS, &svc_params);
}
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_SCREEN_CONTENT_MODE, 0);
}
vpx_codec_control(&codec, VP8E_SET_TOKEN_PARTITIONS, 1);
// This controls the maximum target size of the key frame.
// For generating smaller key frames, use a smaller max_intra_size_pct
// value, like 100 or 200.
{
const int max_intra_size_pct = 1000;
vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT,
max_intra_size_pct);
}
const VpxInterface *decoder = NULL;
decoder = get_vpx_decoder_by_name( "vp9" );
if (!decoder) die("Unsupported codec.");
vpx_codec_ctx_t codec_dec;
vpx_codec_dec_cfg_t cfg_dec;
cfg_dec.w = width;
cfg_dec.h = height;
cfg_dec.threads = 1;
if (vpx_codec_dec_init( &codec_dec, decoder->codec_interface(), &cfg_dec, 0))
die_codec( &codec_dec, "Failed to initialize decoder.");
if(!codec.iface)
die_codec( &codec_dec, "vvv no iface" );
vpx_image_t *img_dec;
frame_avail = 1;
while (frame_avail || got_data) {
struct vpx_usec_timer timer;
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *pkt;
// Update the temporal layer_id. No spatial layers in this test.
layer_id.spatial_layer_id = 0;
layer_id.temporal_layer_id =
cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_codec_control(&codec, VP9E_SET_SVC_LAYER_ID, &layer_id);
} else if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_TEMPORAL_LAYER_ID,
layer_id.temporal_layer_id);
}
flags = layer_flags[frame_cnt % flag_periodicity];
if (layering_mode == 0) flags = 0;
frame_avail = read_frame(&input_ctx, &raw);
if (frame_avail) ++rc.layer_input_frames[layer_id.temporal_layer_id];
vpx_usec_timer_start(&timer);
if (vpx_codec_encode(&codec, frame_avail ? &raw : NULL, pts, 1, flags,
VPX_DL_REALTIME)) {
die_codec(&codec, "Failed to encode frame");
}
vpx_usec_timer_mark(&timer);
cx_time += vpx_usec_timer_elapsed(&timer);
// Reset KF flag.
if (layering_mode != 7) {
layer_flags[0] &= ~VPX_EFLAG_FORCE_KF;
}
got_data = 0;
while ((pkt = vpx_codec_get_cx_data(&codec, &iter))) {
got_data = 1;
switch (pkt->kind) {
case VPX_CODEC_CX_FRAME_PKT:
for (i = cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
i < cfg.ts_number_layers; ++i) {
vpx_video_writer_write_frame(outfile[i], pkt->data.frame.buf,
pkt->data.frame.sz, pts);
++rc.layer_tot_enc_frames[i];
rc.layer_encoding_bitrate[i] += 8.0 * pkt->data.frame.sz;
// Keep count of rate control stats per layer (for non-key frames).
if (i == cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity] &&
!(pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
rc.layer_avg_frame_size[i] += 8.0 * pkt->data.frame.sz;
rc.layer_avg_rate_mismatch[i] +=
fabs(8.0 * pkt->data.frame.sz - rc.layer_pfb[i]) /
rc.layer_pfb[i];
++rc.layer_enc_frames[i];
}
}
// Update for short-time encoding bitrate states, for moving window
// of size rc->window, shifted by rc->window / 2.
// Ignore first window segment, due to key frame.
if (rc.window_size == 0) rc.window_size = 15;
if (frame_cnt > rc.window_size) {
sum_bitrate += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate / rc.window_size) *
(sum_bitrate / rc.window_size);
sum_bitrate = 0.0;
}
}
// Second shifted window.
if (frame_cnt > rc.window_size + rc.window_size / 2) {
sum_bitrate2 += 0.001 * 8.0 * pkt->data.frame.sz * framerate;
if (frame_cnt > 2 * rc.window_size &&
frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
rc.variance_st_encoding_bitrate +=
(sum_bitrate2 / rc.window_size) *
(sum_bitrate2 / rc.window_size);
sum_bitrate2 = 0.0;
}
}
//decode
if( layer_id.temporal_layer_id == 0 ){
vpx_codec_control(&codec_dec, VP9E_SET_SVC_LAYER_ID, &layer_id);
vpx_codec_err_t res = vpx_codec_decode(&codec_dec, pkt->data.frame.buf, pkt->data.frame.sz, NULL, 0);
if( res )
die_codec(&codec_dec, "Failed to decode frame.");
vpx_codec_iter_t iter = NULL;
img_dec = vpx_codec_get_frame(&codec_dec, &iter);
if( img_dec == NULL )
die_codec(&codec_dec, "No frames were decoded.");
if( vpx_codec_get_frame(&codec_dec, &iter) != NULL )
die_codec(&codec_dec, "More than on frame decoded.");
printf( "decoded frame\n" );
}
break;
default: break;
}
}
++frame_cnt;
pts += frame_duration;
}
close_input_file(&input_ctx);
printout_rate_control_summary(&rc, &cfg, frame_cnt);
printf("\n");
printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",
frame_cnt, 1000 * (float)cx_time / (double)(frame_cnt * 1000000),
1000000 * (double)frame_cnt / (double)cx_time);
if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
// Try to rewrite the output file headers with the actual frame count.
for (i = 0; i < cfg.ts_number_layers; ++i) vpx_video_writer_close(outfile[i]);
if (input_ctx.file_type != FILE_TYPE_Y4M) {
vpx_img_free(&raw);
}
#if ROI_MAP
free(roi.roi_map);
#endif
return EXIT_SUCCESS;
}

Why does SChannel TLS limit a message to 32kb

I am working on using SChannel to build a client/server program. One of the things I would like to do is have file sharing. I found some example code of a client program using Schannel to communicate and I am wondering why the max size of a message is 32kb. Here is the example function that does the receiving
int tls_handshake(tls_ctx *c, tls_session *s) {
DWORD flags_in, flags_out;
SecBuffer ib[2], ob[1];
SecBufferDesc in, out;
int len;
// send initial hello
if (!tls_hello(c, s)) {
return 0;
}
flags_in = ISC_REQ_REPLAY_DETECT |
ISC_REQ_CONFIDENTIALITY |
ISC_RET_EXTENDED_ERROR |
ISC_REQ_ALLOCATE_MEMORY |
ISC_REQ_MANUAL_CRED_VALIDATION;
c->ss = SEC_I_CONTINUE_NEEDED;
s->buflen = 0;
while (c->ss == SEC_I_CONTINUE_NEEDED ||
c->ss == SEC_E_INCOMPLETE_MESSAGE ||
c->ss == SEC_I_INCOMPLETE_CREDENTIALS)
{
if (c->ss == SEC_E_INCOMPLETE_MESSAGE)
{
// receive data from server
len = recv(s->sck, &s->buf[s->buflen], s->maxlen - s->buflen, 0);
// socket error?
if (len == SOCKET_ERROR) {
c->ss = SEC_E_INTERNAL_ERROR;
break;
// server disconnected?
} else if (len==0) {
c->ss = SEC_E_INTERNAL_ERROR;
break;
}
// increase buffer position
s->buflen += len;
}
// inspect what we've received
//tls_hex_dump(s->buf, s->buflen);
// input data
ib[0].pvBuffer = s->buf;
ib[0].cbBuffer = s->buflen;
ib[0].BufferType = SECBUFFER_TOKEN;
// empty buffer
ib[1].pvBuffer = NULL;
ib[1].cbBuffer = 0;
ib[1].BufferType = SECBUFFER_VERSION;
in.cBuffers = 2;
in.pBuffers = ib;
in.ulVersion = SECBUFFER_VERSION;
// output from schannel
ob[0].pvBuffer = NULL;
ob[0].cbBuffer = 0;
ob[0].BufferType = SECBUFFER_VERSION;
out.cBuffers = 1;
out.pBuffers = ob;
out.ulVersion = SECBUFFER_VERSION;
c->ss = c->sspi->
InitializeSecurityContextA(
&s->cc, &s->ctx, NULL, flags_in, 0,
SECURITY_NATIVE_DREP, &in, 0, NULL,
&out, &flags_out, NULL);
// what have we got so far?
if (c->ss == SEC_E_OK ||
c->ss == SEC_I_CONTINUE_NEEDED ||
(FAILED(c->ss) && (flags_out & ISC_RET_EXTENDED_ERROR)))
{
// response for server?
if (ob[0].cbBuffer != 0 && ob[0].pvBuffer) {
// send response
tls_send(s->sck, ob[0].pvBuffer, ob[0].cbBuffer);
// free response
c->sspi->FreeContextBuffer(ob[0].pvBuffer);
ob[0].pvBuffer = NULL;
}
}
// incomplete message? continue reading
if (c->ss==SEC_E_INCOMPLETE_MESSAGE) continue;
// completed handshake?
if (c->ss==SEC_E_OK) {
s->established = 1;
// If the "extra" buffer contains data, this is encrypted application
// protocol layer stuff and needs to be saved. The application layer
// will decrypt it later with DecryptMessage.
if (ib[1].BufferType == SECBUFFER_EXTRA) {
DEBUG_PRINT(" [ we have extra data after handshake.\n");
memmove(s->pExtra.pvBuffer,
&s->buf[(s->buflen - ib[1].cbBuffer)], ib[1].cbBuffer);
s->pExtra.cbBuffer = ib[1].cbBuffer;
s->pExtra.BufferType = SECBUFFER_TOKEN;
} else {
// no extra data encountered
s->pExtra.pvBuffer = NULL;
s->pExtra.cbBuffer = 0;
s->pExtra.BufferType = SECBUFFER_EMPTY;
}
break;
}
// some other error
if(FAILED(c->ss)) break;
// Copy any leftover data from the "extra" buffer, and go around again.
if(ib[1].BufferType == SECBUFFER_EXTRA) {
memmove(s->buf, &s->buf[(s->buflen - ib[1].cbBuffer)], ib[1].cbBuffer);
s->buflen = ib[1].cbBuffer;
DEBUG_PRINT(" [ we have %i bytes of extra data.\n", s->buflen);
tls_hex_dump(s->buf, s->buflen);
} else {
s->buflen = 0;
}
}
return c->ss==SEC_E_OK ? 1 : 0;
}
The code comes from a Github I found here: https://github.com/odzhan/shells/blob/master/s6/tls.c
Inside one of his header files he defines
#define TLS_MAX_BUFSIZ 32768
I have also read in other places that this is a limit with TLS. Is it possible to increase that limit? What happens if I need to receive more then that? Like a large file?

Not able to decode mp4 file using latest ffmpeg library : av_decode_video2

I am writing a wrapper code around latest ffmpeg library. I am supplying MP4 files from local system. My problem is that I am unable to get any decoded frames when I use av_decode_video2(). The return value comes out to be negative. I have used av_read_frame() which returns 0. I googled about the problem I am facing but no where could I find the correct explanation. Please give me insight here. Pasting the pseudo code here.
av_init_packet(avpkt);
picture=av_frame_alloc();
pFrameRGB=av_frame_alloc();
codec = avcodec_find_decoder(CODEC_ID_H264);
c= avcodec_alloc_context3(codec)
avcodec_open2(decoderLibraryData->c, decoderLibraryData->codec, NULL)
FormatContext = avformat_alloc_context();
char *pUrl ="./1.MP4";
iRet = avformat_open_input(atContext, pUrl, pFmt, NULL);
if(FormatContext == NULL)
{
printf("could not assign any memory !!!!!!!!! \n");
}
avformat_find_stream_info(FormatContext, NULL);
while(av_read_frame(FormatContext,avpkt) >= 0)
{
len = avcodec_decode_video2(c, picture, &got_picture,avpkt);
printf("CODEC MANAGER len %d Frame decompressed %d \n",len,got_picture);
if (len <= 0)
{
return ERROR;
}
}
}
if(lastHeight != 0 && lastWidth != 0)
{
if(lastWidth != c->width || lastHeight != c->height )
{
av_free(buffer);
buffer = NULL;
lastWidth = c->width;
lastHeight = c->height;
}
}
else
{
lastWidth = c->width;
lastHeight = c->height;
}
decodeFlag = 1;
if(!buffer)
{
int numBytes;
v_mutex_lock(globalCodecLock);
switch(inPixFormat)
{
case RGB:
// Determine required buffer size and allocate buffer
numBytes=avpicture_get_size(PIX_FMT_RGB24, c->width, c->height);
buffer=(uint8_t *)av_malloc(numBytes*sizeof(uint8_t));
avpicture_fill((AVPicture *)pFrameRGB,buffer,PIX_FMT_RGB24,c->width, c->height);
if(cntxt)
sws_freeContext(cntxt);
cntxt = sws_getContext(c->width, c->height, c->pix_fmt,
c->width, c->height, PIX_FMT_RGB24, SWS_BICUBIC, NULL, NULL, NULL);
break;
}
v_mutex_unlock(globalCodecLock);
if(cntxt == NULL)
{
printf("sws_getContext error\n");
return ERROR;
}
}
{
sws_scale(cntxt, picture->data, picture->linesize, 0, c->height, pFrameRGB->data, pFrameRGB->linesize);
if(rgbBuff)
{
if(c->width <= *width && c->height <= *height)
{
saveFrame(pFrameRGB, c->width, c->height, rgbBuff,inPixFormat);
*width = c->width;
*height = c->height;
rs = SUCCESS;
break;
}
else
{
rs = VA_LOWBUFFERSIZE;
}
}
else
{
rs = VA_LOWBUFFERSIZE;
}
}
if(width)
{
*width = c->width;
}
if(height)
{
*height = c->height;
}
if(rs == VA_LOWBUFFERSIZE)
{
break;
}
I am getting the return value of av_read_frame as 0 but av_decode_video2 returns value in negative. I am not able to get any clue here.
Make sure you have called
av_register_all();
or
avcodec_register_all();
at the beginning of your app.
Also it seems the problem is from calling avformat_find_stream_info. Test with the following code:
AVPacket avpkt;
av_init_packet(&avpkt);
AVFrame* picture = av_frame_alloc();
AVFrame* pFrameRGB = av_frame_alloc();
AVFormatContext* c2 = avformat_alloc_context();
char *pUrl = "C:/Sample Videos/20-06-34.MP4";
int video_stream_index = 0;
AVInputFormat* pFmt;
int iRet = avformat_open_input(&c2, pUrl, pFmt, NULL);
AVStream* stream = c2->streams[video_stream_index];
AVCodec* codec = avcodec_find_decoder(stream->codec->codec_id);
avcodec_open2(stream->codec, codec, NULL);
if (c2 == NULL)
{
printf("could not assign any memory !!!!!!!!! \n");
}
while (av_read_frame(c2, &avpkt) >= 0)
{
int got_picture;
int len = avcodec_decode_video2(stream->codec, picture, &got_picture, &avpkt);
printf("CODEC MANAGER len %d Frame decompressed %d \n", len, got_picture);
if (len <= 0)
{
return ERROR;
}
}

FFmpeg C API - syncing video and audio

I am trimming video and having a hard getting the audio to sync correctly. The code below is as close as I've gotten it work. I've tried both re-encoding and not re-encoding the output streams.
The video trims correctly and is written to the output container. The audio stream also trims correctly, but is written to the front of the output container. For example if the trim length is 10s - the correct portion of audio plays for 10s and then the correct portion of video plays.
//////// audio stream ////////
const AVStream *input_stream_audio = input_container->streams[audio_stream_index];
const AVCodec *decoder_audio = avcodec_find_decoder(input_stream_audio->codec->codec_id);
if(!decoder_audio) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Audio decoder not found");
return -1;
}
if(avcodec_open2(input_stream_audio->codec, decoder_audio, NULL) < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error opening audio decoder");
return -1;
}
AVStream *output_stream_audio = avformat_new_stream(output_container, NULL);
if(avcodec_copy_context(output_stream_audio->codec, input_stream_audio->codec) != 0){
LOGE("=> Failed to Copy audio Context ");
return -1;
}
else {
LOGI("=> Copied audio context ");
output_stream_audio->codec->codec_id = input_stream_audio->codec->codec_id;
output_stream_audio->codec->codec_tag = 0;
output_stream_audio->pts = input_stream_audio->pts;
output_stream_audio->time_base.num = input_stream_audio->time_base.num;
output_stream_audio->time_base.den = input_stream_audio->time_base.den;
}
if(avio_open(&output_container->pb, output_file, AVIO_FLAG_WRITE) < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error opening output file");
return -1;
}
// allocate frame for conversion
decoded_frame = avcodec_alloc_frame();
if(!decoded_frame) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error allocating frame");
return -1;
}
av_dump_format(input_container, 0, input_file, 0);
avformat_write_header(output_container, NULL);
av_init_packet(&decoded_packet);
decoded_packet.data = NULL;
decoded_packet.size = 0;
int current_frame_num = 1;
int current_frame_num_audio = 1;
int got_frame, len;
AVRational default_timebase;
default_timebase.num = 1;
default_timebase.den = AV_TIME_BASE;
int64_t starttime_int64 = av_rescale_q((int64_t)( 12.0 * AV_TIME_BASE ), AV_TIME_BASE_Q, input_stream->time_base);
int64_t endtime_int64 = av_rescale_q((int64_t)( 18.0 * AV_TIME_BASE ), AV_TIME_BASE_Q, input_stream->time_base);
LOGI("=> starttime_int64: %" PRId64, starttime_int64);
LOGI("=> endtime_int64: %" PRId64, endtime_int64);
int64_t starttime_int64_audio = av_rescale_q((int64_t)( 12.0 * AV_TIME_BASE ), AV_TIME_BASE_Q, input_stream_audio->time_base);
int64_t endtime_int64_audio = av_rescale_q((int64_t)( 18.0 * AV_TIME_BASE ), AV_TIME_BASE_Q, input_stream_audio->time_base);
LOGI("=> starttime_int64_audio: %" PRId64, starttime_int64_audio);
LOGI("=> endtime_int64_audio: %" PRId64, endtime_int64_audio);
// loop input container and decode frames
while(av_read_frame(input_container, &decoded_packet)>=0) {
// video packets
if (decoded_packet.stream_index == video_stream_index) {
len = avcodec_decode_video2(input_stream->codec, decoded_frame, &got_frame, &decoded_packet);
if(len < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> No frames to decode");
return -1;
}
// this is the trim range we're looking for
if(got_frame && decoded_frame->pkt_pts >= starttime_int64 && decoded_frame->pkt_pts <= endtime_int64) {
av_init_packet(&encoded_packet);
encoded_packet.data = NULL;
encoded_packet.size = 0;
ret = avcodec_encode_video2(output_stream->codec, &encoded_packet, decoded_frame, &got_frame);
if (ret < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error encoding frames");
return ret;
}
if(got_frame) {
if (output_stream->codec->coded_frame->key_frame) {
encoded_packet.flags |= AV_PKT_FLAG_KEY;
}
encoded_packet.stream_index = output_stream->index;
encoded_packet.pts = av_rescale_q(current_frame_num, output_stream->codec->time_base, output_stream->time_base);
encoded_packet.dts = av_rescale_q(current_frame_num, output_stream->codec->time_base, output_stream->time_base);
ret = av_interleaved_write_frame(output_container, &encoded_packet);
if (ret < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error encoding frames");
return ret;
}
else {
current_frame_num +=1;
}
}
av_free_packet(&encoded_packet);
}
}
// audio packets
else if(decoded_packet.stream_index == audio_stream_index) {
// this is the trim range we're looking for
if(decoded_packet.pts >= starttime_int64_audio && decoded_packet.pts <= endtime_int64_audio) {
av_init_packet(&encoded_packet);
encoded_packet.data = decoded_packet.data;
encoded_packet.size = decoded_packet.size;
encoded_packet.stream_index = audio_stream_index;
encoded_packet.pts = av_rescale_q(current_frame_num_audio, output_stream_audio->codec->time_base, output_stream_audio->time_base);
encoded_packet.dts = av_rescale_q(current_frame_num_audio, output_stream_audio->codec->time_base, output_stream_audio->time_base);
ret = av_interleaved_write_frame(output_container, &encoded_packet);
if (ret < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error encoding frames");
return ret;
}
else {
current_frame_num_audio +=1;
}
av_free_packet(&encoded_packet);
}
}
}
Edit
I have slight improvement on the initial code. The audio and video are still not perfectly synced, but the original problem of the audio playing first followed by the video is resolved.
I'm now writing the decoded packet to the output container rather than re-encoding it.
In the end though I have the same problem - the trimmed video's audio and video streams are not perfectly synced.
// audio packets
else if(decoded_packet.stream_index == audio_stream_index) {
// this is the trim range we're looking for
if(decoded_packet.pts >= starttime_int64_audio && decoded_packet.pts <= endtime_int64_audio) {
ret = av_interleaved_write_frame(output_container, &decoded_packet);
if (ret < 0) {
cleanup(decoded_packet, output_container, decoded_frame);
avformat_close_input(&input_container);
LOGE("=> Error writing audio frame (%s)", av_err2str(ret));
return ret;
}
else {
current_frame_num_audio +=1;
}
}
else if(decoded_frame->pkt_pts > endtime_int64_audio) {
audio_copy_complete = true;
}
}
I believe you should be able to make this work if you use correctly set up both codec context and stream time bases, and then after encode_video2 and encode_audio2 you call av_packet_rescale_tb with such timebases.

How to cut video with FFmpeg C API

How can I cut video with FFmpeg C API? From 00:10:00 to 00:20:00 for example.
What functions I need to use?
I convert video using this function:
int convert(char *file) {
AVFrame *frame;
AVPacket inPacket, outPacket;
if(avio_open(&outFormatContext->pb, file, AVIO_FLAG_WRITE) < 0) {
fprintf(stderr, "convert(): cannot open out file\n");
return 0;
}
avformat_write_header(outFormatContext, NULL);
frame = avcodec_alloc_frame();
av_init_packet(&inPacket);
while(av_read_frame(inFormatContext, &inPacket) >= 0) {
if(inPacket.stream_index == inVideoStreamIndex) {
avcodec_decode_video2(inCodecContext, frame, &frameFinished, &inPacket);
if(frameFinished) {
av_init_packet(&outPacket);
avcodec_encode_video2(outCodecContext, &outPacket, frame, &outputed);
if(outputed) {
if (av_write_frame(outFormatContext, &outPacket) != 0) {
fprintf(stderr, "convert(): error while writing video frame\n");
return 0;
}
}
av_free_packet(&outPacket);
}
}
}
av_write_trailer(outFormatContext);
av_free_packet(&inPacket);
return 1;
}
As Wagner Patriota says, "if you just wanna cut the video, you don't need to reencode the video if you want". Here is the code based on ffmpeg remuxing.c example that you don't need to reencode the video.
#include <libavutil/timestamp.h>
#include <libavformat/avformat.h>
static void log_packet(const AVFormatContext *fmt_ctx, const AVPacket *pkt, const char *tag)
{
AVRational *time_base = &fmt_ctx->streams[pkt->stream_index]->time_base;
printf("%s: pts:%s pts_time:%s dts:%s dts_time:%s duration:%s duration_time:%s stream_index:%d\n",
tag,
av_ts2str(pkt->pts), av_ts2timestr(pkt->pts, time_base),
av_ts2str(pkt->dts), av_ts2timestr(pkt->dts, time_base),
av_ts2str(pkt->duration), av_ts2timestr(pkt->duration, time_base),
pkt->stream_index);
}
int cut_video(double from_seconds, double end_seconds, const char* in_filename, const char* out_filename) {
AVOutputFormat *ofmt = NULL;
AVFormatContext *ifmt_ctx = NULL, *ofmt_ctx = NULL;
AVPacket pkt;
int ret, i;
av_register_all();
if ((ret = avformat_open_input(&ifmt_ctx, in_filename, 0, 0)) < 0) {
fprintf(stderr, "Could not open input file '%s'", in_filename);
goto end;
}
if ((ret = avformat_find_stream_info(ifmt_ctx, 0)) < 0) {
fprintf(stderr, "Failed to retrieve input stream information");
goto end;
}
av_dump_format(ifmt_ctx, 0, in_filename, 0);
avformat_alloc_output_context2(&ofmt_ctx, NULL, NULL, out_filename);
if (!ofmt_ctx) {
fprintf(stderr, "Could not create output context\n");
ret = AVERROR_UNKNOWN;
goto end;
}
ofmt = ofmt_ctx->oformat;
for (i = 0; i < ifmt_ctx->nb_streams; i++) {
AVStream *in_stream = ifmt_ctx->streams[i];
AVStream *out_stream = avformat_new_stream(ofmt_ctx, in_stream->codec->codec);
if (!out_stream) {
fprintf(stderr, "Failed allocating output stream\n");
ret = AVERROR_UNKNOWN;
goto end;
}
ret = avcodec_copy_context(out_stream->codec, in_stream->codec);
if (ret < 0) {
fprintf(stderr, "Failed to copy context from input to output stream codec context\n");
goto end;
}
out_stream->codec->codec_tag = 0;
if (ofmt_ctx->oformat->flags & AVFMT_GLOBALHEADER)
out_stream->codec->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
}
av_dump_format(ofmt_ctx, 0, out_filename, 1);
if (!(ofmt->flags & AVFMT_NOFILE)) {
ret = avio_open(&ofmt_ctx->pb, out_filename, AVIO_FLAG_WRITE);
if (ret < 0) {
fprintf(stderr, "Could not open output file '%s'", out_filename);
goto end;
}
}
ret = avformat_write_header(ofmt_ctx, NULL);
if (ret < 0) {
fprintf(stderr, "Error occurred when opening output file\n");
goto end;
}
// int indexs[8] = {0};
// int64_t start_from = 8*AV_TIME_BASE;
ret = av_seek_frame(ifmt_ctx, -1, from_seconds*AV_TIME_BASE, AVSEEK_FLAG_ANY);
if (ret < 0) {
fprintf(stderr, "Error seek\n");
goto end;
}
int64_t *dts_start_from = malloc(sizeof(int64_t) * ifmt_ctx->nb_streams);
memset(dts_start_from, 0, sizeof(int64_t) * ifmt_ctx->nb_streams);
int64_t *pts_start_from = malloc(sizeof(int64_t) * ifmt_ctx->nb_streams);
memset(pts_start_from, 0, sizeof(int64_t) * ifmt_ctx->nb_streams);
while (1) {
AVStream *in_stream, *out_stream;
ret = av_read_frame(ifmt_ctx, &pkt);
if (ret < 0)
break;
in_stream = ifmt_ctx->streams[pkt.stream_index];
out_stream = ofmt_ctx->streams[pkt.stream_index];
log_packet(ifmt_ctx, &pkt, "in");
if (av_q2d(in_stream->time_base) * pkt.pts > end_seconds) {
av_free_packet(&pkt);
break;
}
if (dts_start_from[pkt.stream_index] == 0) {
dts_start_from[pkt.stream_index] = pkt.dts;
printf("dts_start_from: %s\n", av_ts2str(dts_start_from[pkt.stream_index]));
}
if (pts_start_from[pkt.stream_index] == 0) {
pts_start_from[pkt.stream_index] = pkt.pts;
printf("pts_start_from: %s\n", av_ts2str(pts_start_from[pkt.stream_index]));
}
/* copy packet */
pkt.pts = av_rescale_q_rnd(pkt.pts - pts_start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
pkt.dts = av_rescale_q_rnd(pkt.dts - dts_start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
if (pkt.pts < 0) {
pkt.pts = 0;
}
if (pkt.dts < 0) {
pkt.dts = 0;
}
pkt.duration = (int)av_rescale_q((int64_t)pkt.duration, in_stream->time_base, out_stream->time_base);
pkt.pos = -1;
log_packet(ofmt_ctx, &pkt, "out");
printf("\n");
ret = av_interleaved_write_frame(ofmt_ctx, &pkt);
if (ret < 0) {
fprintf(stderr, "Error muxing packet\n");
break;
}
av_free_packet(&pkt);
}
free(dts_start_from);
free(pts_start_from);
av_write_trailer(ofmt_ctx);
end:
avformat_close_input(&ifmt_ctx);
/* close output */
if (ofmt_ctx && !(ofmt->flags & AVFMT_NOFILE))
avio_closep(&ofmt_ctx->pb);
avformat_free_context(ofmt_ctx);
if (ret < 0 && ret != AVERROR_EOF) {
fprintf(stderr, "Error occurred: %s\n", av_err2str(ret));
return 1;
}
return 0;
}
If you just wanna cut the video, you don't need to reencode the video if you want.
So I am supposing you wanna cut and reencode for some reason. So, based on your code:
Observe you must to have access to the video AVStream* structure... I named it as inVideoStream.
int convert_and_cut(char *file, float starttime, float endtime) {
AVFrame *frame;
AVPacket inPacket, outPacket;
if(avio_open(&outFormatContext->pb, file, AVIO_FLAG_WRITE) < 0) {
fprintf(stderr, "convert(): cannot open out file\n");
return 0;
}
// seek to the start time you wish.
// BEGIN
AVRational default_timebase;
default_timebase.num = 1;
default_timebase.den = AV_TIME_BASE;
// suppose you have access to the "inVideoStream" of course
int64_t starttime_int64 = av_rescale_q((int64_t)( starttime * AV_TIME_BASE ), default_timebase, inVideoStream->time_base);
int64_t endtime_int64 = av_rescale_q((int64_t)( endtime * AV_TIME_BASE ), default_timebase, inVideoStream->time_base);
if(avformat_seek_file(inFormatContext, inVideoStreamIndex, INT64_MIN, starttime_int64, INT64_MAX, 0) < 0) {
// error... do something...
return 0; // usually 0 is used for success in C, but I am following your code.
}
avcodec_flush_buffers( inVideoStream->codec );
// END
avformat_write_header(outFormatContext, NULL);
frame = avcodec_alloc_frame();
av_init_packet(&inPacket);
// you used avformat_seek_file() to seek CLOSE to the point you want... in order to give precision to your seek,
// just go on reading the packets and checking the packets PTS (presentation timestamp)
while(av_read_frame(inFormatContext, &inPacket) >= 0) {
if(inPacket.stream_index == inVideoStreamIndex) {
avcodec_decode_video2(inCodecContext, frame, &frameFinished, &inPacket);
// this line guarantees you are getting what you really want.
if(frameFinished && frame->pkt_pts >= starttime_int64 && frame->pkt_pts <= endtime_int64) {
av_init_packet(&outPacket);
avcodec_encode_video2(outCodecContext, &outPacket, frame, &outputed);
if(outputed) {
if (av_write_frame(outFormatContext, &outPacket) != 0) {
fprintf(stderr, "convert(): error while writing video frame\n");
return 0;
}
}
av_free_packet(&outPacket);
}
// exit the loop if you got the frames you want.
if(frame->pkt_pts > endtime_int64) {
break;
}
}
}
av_write_trailer(outFormatContext);
av_free_packet(&inPacket);
return 1;
}
In addition, as ustin says, the code based on ffmpeg remuxing.c example.
add out_stream->time_base = in_stream->time_base; before out_stream->codec->codec_tag = 0;
add int64_t *start_from = NULL; after int stream_mapping_size = 0;
add start_from = av_mallocz_array(stream_mapping_size, sizeof(*start_from)); after stream_mapping = av_mallocz_array(stream_mapping_size, sizeof(*stream_mapping));
add start_from[stream_mapping[i]] = INT16_MIN; after stream_mapping[i] = stream_index++;
add
if (start_from[pkt.stream_index] == INT16_MIN) {
int64_t dts = pkt.dts;
int64_t pts = pkt.pts;
int64_t min_ts = dts > pts ? pts : dts;
start_from[pkt.stream_index] = min_ts < 0 ? 0 : min_ts;
}
After
if (pkt.stream_index >= stream_mapping_size ||
stream_mapping[pkt.stream_index] < 0) {
av_packet_unref(&pkt);
continue;
}
change justin code, about pts/dst set value to this:
pkt.pts = av_rescale_q_rnd(pkt.pts - start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
pkt.dts = av_rescale_q_rnd(pkt.dts - start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
add free(start_from); before end:
Tips: start/end timestamp like justin says

Resources