static int query_formats(AVFilterGraph *graph, AVClass *log_ctx)
{
int i, j, ret;
char filt_args[128];
AVFilterFormats *formats;
AVFilterChannelLayouts *chlayouts;
AVFilterFormats *samplerates;
int scaler_count = 0, resampler_count = 0;
for (j = 0; j < 2; j++) {
/* ask all the sub-filters for their supported media formats */
for (i = 0; i < graph->filter_count; i++) {
/* Call query_formats on sources first.
This is a temporary workaround for amerge,
until format renegociation is implemented. */
if (!graph->filters[i]->nb_inputs == j)
continue;
if (graph->filters[i]->filter->query_formats)
ret = filter_query_formats(graph->filters[i]);
else
ret = ff_default_query_formats(graph->filters[i]);
if (ret < 0)
return ret;
}
}
/* go through and merge as many format lists as possible */
for (i = 0; i < graph->filter_count; i++) {
AVFilterContext *filter = graph->filters[i];
for (j = 0; j < filter->nb_inputs; j++) {
AVFilterLink *link = filter->inputs[j];
#if 0
if (!link) continue;
if (!link->in_formats || !link->out_formats)
return AVERROR(EINVAL);
if (link->type == AVMEDIA_TYPE_VIDEO &&
!ff_merge_formats(link->in_formats, link->out_formats)) {
/* couldn't merge format lists, auto-insert scale filter */
snprintf(filt_args, sizeof(filt_args), "0:0:%s",
graph->scale_sws_opts);
if (ret = insert_conv_filter(graph, link, "scale", filt_args))
return ret;
}
else if (link->type == AVMEDIA_TYPE_AUDIO) {
if (!link->in_channel_layouts || !link->out_channel_layouts)
return AVERROR(EINVAL);
/* Merge all three list before checking: that way, in all
* three categories, aconvert will use a common format
* whenever possible. */
formats = ff_merge_formats(link->in_formats, link->out_formats);
chlayouts = ff_merge_channel_layouts(link->in_channel_layouts , link->out_channel_layouts);
samplerates = ff_merge_samplerates (link->in_samplerates, link->out_samplerates);
if (!formats || !chlayouts || !samplerates)
if (ret = insert_conv_filter(graph, link, "aresample", NULL))
return ret;
#else
int convert_needed = 0;
if (!link)
continue;
if (link->in_formats != link->out_formats &&
!ff_merge_formats(link->in_formats,
link->out_formats))
convert_needed = 1;
if (link->type == AVMEDIA_TYPE_AUDIO) {
if (link->in_channel_layouts != link->out_channel_layouts &&
!ff_merge_channel_layouts(link->in_channel_layouts,
link->out_channel_layouts))
convert_needed = 1;
if (link->in_samplerates != link->out_samplerates &&
!ff_merge_samplerates(link->in_samplerates,
link->out_samplerates))
convert_needed = 1;
}
if (convert_needed) {
AVFilterContext *convert;
AVFilter *filter;
AVFilterLink *inlink, *outlink;
char scale_args[256];
char inst_name[30];
/* couldn't merge format lists. auto-insert conversion filter */
switch (link->type) {
case AVMEDIA_TYPE_VIDEO:
if (!(filter = avfilter_get_by_name("scale"))) {
av_log(log_ctx, AV_LOG_ERROR, "'scale' filter "
"not present, cannot convert pixel formats.\n");
return AVERROR(EINVAL);
}
snprintf(inst_name, sizeof(inst_name), "auto-inserted scaler %d",
scaler_count++);
snprintf(scale_args, sizeof(scale_args), "0:0:%s", graph->scale_sws_opts);
if ((ret = avfilter_graph_create_filter(&convert, filter,
inst_name, scale_args, NULL,
graph)) < 0)
return ret;
break;
case AVMEDIA_TYPE_AUDIO:
if (!(filter = avfilter_get_by_name("aresample"))) {
av_log(log_ctx, AV_LOG_ERROR, "'aresample' filter "
"not present, cannot convert audio formats.\n");
return AVERROR(EINVAL);
}
snprintf(inst_name, sizeof(inst_name), "auto-inserted resampler %d",
resampler_count++);
if ((ret = avfilter_graph_create_filter(&convert, filter,
inst_name, NULL, NULL, graph)) < 0)
return ret;
break;
default:
return AVERROR(EINVAL);
}
if ((ret = avfilter_insert_filter(link, convert, 0, 0)) < 0)
return ret;
filter_query_formats(convert);
inlink = convert->inputs[0];
outlink = convert->outputs[0];
if (!ff_merge_formats( inlink->in_formats, inlink->out_formats) ||
!ff_merge_formats(outlink->in_formats, outlink->out_formats))
ret |= AVERROR(ENOSYS);
if (inlink->type == AVMEDIA_TYPE_AUDIO &&
(!ff_merge_samplerates(inlink->in_samplerates,
inlink->out_samplerates) ||
!ff_merge_channel_layouts(inlink->in_channel_layouts,
inlink->out_channel_layouts)))
ret |= AVERROR(ENOSYS);
if (outlink->type == AVMEDIA_TYPE_AUDIO &&
(!ff_merge_samplerates(outlink->in_samplerates,
outlink->out_samplerates) ||
!ff_merge_channel_layouts(outlink->in_channel_layouts,
outlink->out_channel_layouts)))
ret |= AVERROR(ENOSYS);
if (ret < 0) {
av_log(log_ctx, AV_LOG_ERROR,
"Impossible to convert between the formats supported by the filter "
"'%s' and the filter '%s'\n", link->src->name, link->dst->name);
return ret;
}
#endif
}
}
}
return 0;
}
static int pick_format(AVFilterLink *link, AVFilterLink *ref)
{
if (!link || !link->in_formats)
return 0;
if (link->type == AVMEDIA_TYPE_VIDEO) {
if(ref && ref->type == AVMEDIA_TYPE_VIDEO){
int has_alpha= av_pix_fmt_descriptors[ref->format].nb_components % 2 == 0;
enum PixelFormat best= PIX_FMT_NONE;
int i;
for (i=0; i<link->in_formats->format_count; i++) {
enum PixelFormat p = link->in_formats->formats[i];
best= avcodec_find_best_pix_fmt2(best, p, ref->format, has_alpha, NULL);
}
av_log(link->src,AV_LOG_DEBUG, "picking %s out of %d ref:%s alpha:%d\n",
av_get_pix_fmt_name(best), link->in_formats->format_count,
av_get_pix_fmt_name(ref->format), has_alpha);
link->in_formats->formats[0] = best;
}
}
link->in_formats->format_count = 1;
link->format = link->in_formats->formats[0];
if (link->type == AVMEDIA_TYPE_AUDIO) {
if (!link->in_samplerates->format_count) {
av_log(link->src, AV_LOG_ERROR, "Cannot select sample rate for"
" the link between filters %s and %s.\n", link->src->name,
link->dst->name);
return AVERROR(EINVAL);
}
link->in_samplerates->format_count = 1;
link->sample_rate = link->in_samplerates->formats[0];
if (!link->in_channel_layouts->nb_channel_layouts) {
av_log(link->src, AV_LOG_ERROR, "Cannot select channel layout for"
"the link between filters %s and %s.\n", link->src->name,
link->dst->name);
return AVERROR(EINVAL);
}
link->in_channel_layouts->nb_channel_layouts = 1;
link->channel_layout = link->in_channel_layouts->channel_layouts[0];
}
ff_formats_unref(&link->in_formats);
ff_formats_unref(&link->out_formats);
ff_formats_unref(&link->in_samplerates);
ff_formats_unref(&link->out_samplerates);
ff_channel_layouts_unref(&link->in_channel_layouts);
ff_channel_layouts_unref(&link->out_channel_layouts);
return 0;
}
#define REDUCE_FORMATS(fmt_type, list_type, list, var, nb, add_format) \
do { \
for (i = 0; i < filter->nb_inputs; i++) { \
AVFilterLink *link = filter->inputs[i]; \
fmt_type fmt; \
\
if (!link->out_ ## list || link->out_ ## list->nb != 1) \
continue; \
fmt = link->out_ ## list->var[0]; \
\
for (j = 0; j < filter->nb_outputs; j++) { \
AVFilterLink *out_link = filter->outputs[j]; \
list_type *fmts; \
\
if (link->type != out_link->type || \
out_link->in_ ## list->nb == 1) \
continue; \
fmts = out_link->in_ ## list; \
\
if (!out_link->in_ ## list->nb) { \
add_format(&out_link->in_ ##list, fmt); \
break; \
} \
\
for (k = 0; k < out_link->in_ ## list->nb; k++) \
if (fmts->var[k] == fmt) { \
fmts->var[0] = fmt; \
fmts->nb = 1; \
ret = 1; \
break; \
} \
} \
} \
} while (0)
static int reduce_formats_on_filter(AVFilterContext *filter)
{
int i, j, k, ret = 0;
REDUCE_FORMATS(int, AVFilterFormats, formats, formats,
format_count, ff_add_format);
REDUCE_FORMATS(int, AVFilterFormats, samplerates, formats,
format_count, ff_add_format);
REDUCE_FORMATS(uint64_t, AVFilterChannelLayouts, channel_layouts,
channel_layouts, nb_channel_layouts, ff_add_channel_layout);
return ret;
}
MediaRet MediaRecorder::setup_sound_stream(const char *fname, AVOutputFormat *fmt)
{
oc = avformat_alloc_context();
if(!oc)
return MRET_ERR_NOMEM;
oc->oformat = fmt;
strncpy(oc->filename, fname, sizeof(oc->filename) - 1);
oc->filename[sizeof(oc->filename) - 1] = 0;
if(fmt->audio_codec == CODEC_ID_NONE)
return MRET_OK;
AVCodecContext *ctx;
#if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(53,10,0)
aud_st = av_new_stream(oc, 1);
#else
aud_st = avformat_new_stream(oc, NULL);
#endif
if(!aud_st) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
AVCodec *codec = avcodec_find_encoder(fmt->audio_codec);
ctx = aud_st->codec;
ctx->codec_id = fmt->audio_codec;
ctx->codec_type = AVMEDIA_TYPE_AUDIO;
// Some encoders don't like int16_t (SAMPLE_FMT_S16)
ctx->sample_fmt = codec->sample_fmts[0];
// This was changed in the initial ffmpeg 3.0 update,
// but shouldn't (as far as I'm aware) cause problems with older versions
ctx->bit_rate = 128000; // arbitrary; in case we're generating mp3
ctx->sample_rate = soundGetSampleRate();
ctx->channels = 2;
ctx->time_base.den = 60;
ctx->time_base.num = 1;
if(fmt->flags & AVFMT_GLOBALHEADER)
ctx->flags |= CODEC_FLAG_GLOBAL_HEADER;
#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(53,6,0)
if(!codec || avcodec_open(ctx, codec)) {
#else
if(!codec || avcodec_open2(ctx, codec, NULL)) {
#endif
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOCODEC;
}
return MRET_OK;
}
MediaRet MediaRecorder::setup_video_stream(const char *fname, int w, int h, int d)
{
AVCodecContext *ctx;
#if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(53,10,0)
vid_st = av_new_stream(oc, 0);
#else
vid_st = avformat_new_stream(oc, NULL);
#endif
if(!vid_st) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
ctx = vid_st->codec;
ctx->codec_id = oc->oformat->video_codec;
ctx->codec_type = AVMEDIA_TYPE_VIDEO;
ctx->width = w;
ctx->height = h;
ctx->time_base.den = 60;
ctx->time_base.num = 1;
// dunno if any of these help; some output just looks plain crappy
// will have to investigate further
ctx->bit_rate = 400000;
ctx->gop_size = 12;
ctx->max_b_frames = 2;
switch(d) {
case 16:
// FIXME: test & make endian-neutral
pixfmt = PIX_FMT_RGB565LE;
break;
case 24:
pixfmt = PIX_FMT_RGB24;
break;
case 32:
default: // should never be anything else
pixfmt = PIX_FMT_RGBA;
break;
}
ctx->pix_fmt = pixfmt;
pixsize = d >> 3;
linesize = pixsize * w;
ctx->max_b_frames = 2;
if(oc->oformat->flags & AVFMT_GLOBALHEADER)
ctx->flags |= CODEC_FLAG_GLOBAL_HEADER;
AVCodec *codec = avcodec_find_encoder(oc->oformat->video_codec);
// make sure RGB is supported (mostly not)
if(codec->pix_fmts) {
const enum PixelFormat *p;
#if LIBAVCODEC_VERSION_MAJOR < 55
int64_t mask = 0;
#endif
for(p = codec->pix_fmts; *p != -1; p++) {
// may get complaints about 1LL; thus the cast
#if LIBAVCODEC_VERSION_MAJOR < 55
mask |= ((int64_t)1) << *p;
#endif
if(*p == pixfmt)
break;
}
if(*p == -1) {
// if not supported, use a converter to the next best format
// this is swscale, the converter used by the output demo
#if LIBAVCODEC_VERSION_MAJOR < 55
enum PixelFormat dp = (PixelFormat)avcodec_find_best_pix_fmt(mask, pixfmt, 0, NULL);
#else
#if LIBAVCODEC_VERSION_MICRO >= 100
// FFmpeg
enum AVPixelFormat dp = avcodec_find_best_pix_fmt_of_list(codec->pix_fmts, pixfmt, 0, NULL);
#else
// Libav
enum AVPixelFormat dp = avcodec_find_best_pix_fmt2(codec->pix_fmts, pixfmt, 0, NULL);
#endif
#endif
if(dp == -1)
dp = codec->pix_fmts[0];
if(!(convpic = avcodec_alloc_frame()) ||
avpicture_alloc((AVPicture *)convpic, dp, w, h) < 0) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
#if LIBSWSCALE_VERSION_INT < AV_VERSION_INT(0, 12, 0)
converter = sws_getContext(w, h, pixfmt, w, h, dp, SWS_BICUBIC,
NULL, NULL, NULL);
#else
converter = sws_alloc_context();
// what a convoluted, inefficient way to set options
av_opt_set_int(converter, "sws_flags", SWS_BICUBIC, 0);
av_opt_set_int(converter, "srcw", w, 0);
av_opt_set_int(converter, "srch", h, 0);
av_opt_set_int(converter, "dstw", w, 0);
av_opt_set_int(converter, "dsth", h, 0);
av_opt_set_int(converter, "src_format", pixfmt, 0);
av_opt_set_int(converter, "dst_format", dp, 0);
sws_init_context(converter, NULL, NULL);
#endif
ctx->pix_fmt = dp;
}
}
#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(53,6,0)
if(!codec || avcodec_open(ctx, codec)) {
#else
if(!codec || avcodec_open2(ctx, codec, NULL)) {
#endif
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOCODEC;
}
return MRET_OK;
}
MediaRet MediaRecorder::finish_setup(const char *fname)
{
if(audio_buf)
free(audio_buf);
if(audio_buf2)
free(audio_buf2);
audio_buf2 = NULL;
in_audio_buf2 = 0;
if(aud_st) {
frame_len = aud_st->codec->frame_size * 4;
sample_len = soundGetSampleRate() * 4 / 60;
switch(aud_st->codec->codec_id) {
case CODEC_ID_PCM_S16LE:
case CODEC_ID_PCM_S16BE:
case CODEC_ID_PCM_U16LE:
case CODEC_ID_PCM_U16BE:
frame_len = sample_len;
}
audio_buf = (uint8_t *)malloc(AUDIO_BUF_LEN);
if(!audio_buf) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
if(frame_len != sample_len && (frame_len > sample_len || sample_len % frame_len)) {
audio_buf2 = (uint16_t *)malloc(frame_len);
if(!audio_buf2) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
}
} else
audio_buf = NULL;
if(video_buf)
free(video_buf);
if(vid_st) {
video_buf = (uint8_t *)malloc(VIDEO_BUF_LEN);
if(!video_buf) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_NOMEM;
}
} else {
video_buf = NULL;
}
if(!(oc->oformat->flags & AVFMT_NOFILE)) {
if(avio_open(&oc->pb, fname, AVIO_FLAG_WRITE) < 0) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_FERR;
}
}
avformat_write_header(oc, NULL);
return MRET_OK;
}
MediaRet MediaRecorder::Record(const char *fname, int width, int height, int depth)
{
if(oc)
return MRET_ERR_RECORDING;
aud_st = vid_st = NULL;
AVOutputFormat *fmt = av_guess_format(NULL, fname, NULL);
if(!fmt)
fmt = av_guess_format("avi", NULL, NULL);
if(!fmt || fmt->video_codec == CODEC_ID_NONE)
return MRET_ERR_FMTGUESS;
MediaRet ret;
if((ret = setup_sound_stream(fname, fmt)) == MRET_OK &&
(ret = setup_video_stream(fname, width, height, depth)) == MRET_OK)
ret = finish_setup(fname);
return ret;
}
MediaRet MediaRecorder::Record(const char *fname)
{
if(oc)
return MRET_ERR_RECORDING;
aud_st = vid_st = NULL;
AVOutputFormat *fmt = av_guess_format(NULL, fname, NULL);
if(!fmt)
fmt = av_guess_format("wav", NULL, NULL);
if(!fmt || fmt->audio_codec == CODEC_ID_NONE)
return MRET_ERR_FMTGUESS;
MediaRet ret;
if((ret = setup_sound_stream(fname, fmt)) == MRET_OK)
ret = finish_setup(fname);
return ret;
}
void MediaRecorder::Stop()
{
if(oc) {
if(in_audio_buf2)
AddFrame((uint16_t *)0);
av_write_trailer(oc);
avformat_free_context(oc);
oc = NULL;
}
if(audio_buf) {
free(audio_buf);
audio_buf = NULL;
}
if(video_buf) {
free(video_buf);
video_buf = NULL;
}
if(audio_buf2) {
free(audio_buf2);
audio_buf2 = NULL;
}
if(convpic) {
avpicture_free((AVPicture *)convpic);
av_free(convpic);
convpic = NULL;
}
if(converter) {
sws_freeContext(converter);
converter = NULL;
}
}
MediaRecorder::~MediaRecorder()
{
Stop();
}
// Still needs updating for avcodec_encode_video2
MediaRet MediaRecorder::AddFrame(const uint8_t *vid)
{
if(!oc || !vid_st)
return MRET_OK;
AVCodecContext *ctx = vid_st->codec;
AVPacket pkt;
#if LIBAVCODEC_VERSION_MAJOR > 56
int ret, got_packet = 0;
#endif
// strip borders. inconsistent between depths for some reason
// but fortunately consistent between gb/gba.
int tbord, rbord;
switch(pixsize) {
case 2:
// 16-bit: 2 @ right, 1 @ top
tbord = 1; rbord = 2; break;
case 3:
// 24-bit: no border
tbord = rbord = 0; break;
case 4:
// 32-bit: 1 @ right, 1 @ top
tbord = 1; rbord = 1; break;
}
avpicture_fill((AVPicture *)pic, (uint8_t *)vid + tbord * (linesize + pixsize * rbord),
(PixelFormat)pixfmt, ctx->width + rbord, ctx->height);
// satisfy stupid sws_scale()'s integrity check
pic->data[1] = pic->data[2] = pic->data[3] = pic->data[0];
pic->linesize[1] = pic->linesize[2] = pic->linesize[3] = pic->linesize[0];
AVFrame *f = pic;
if(converter) {
sws_scale(converter, pic->data, pic->linesize, 0, ctx->height,
convpic->data, convpic->linesize);
f = convpic;
}
av_init_packet(&pkt);
pkt.stream_index = vid_st->index;
if(oc->oformat->flags & AVFMT_RAWPICTURE) {
// this won't work due to border
// not sure what formats set this, anyway
pkt.flags |= AV_PKT_FLAG_KEY;
pkt.data = f->data[0];
pkt.size = linesize * ctx->height;
} else {
#if LIBAVCODEC_VERSION_MAJOR > 56
pkt.data = video_buf;
pkt.size = VIDEO_BUF_LEN;
f->format = ctx->pix_fmt;
f->width = ctx->width;
f->height = ctx->height;
ret = avcodec_encode_video2(ctx, &pkt, f, &got_packet);
if(!ret && got_packet && ctx->coded_frame) {
ctx->coded_frame->pts = pkt.pts;
ctx->coded_frame->key_frame = !!(pkt.flags & AV_PKT_FLAG_KEY);
}
#else
pkt.size = avcodec_encode_video(ctx, video_buf, VIDEO_BUF_LEN, f);
#endif
if(!pkt.size)
return MRET_OK;
if(ctx->coded_frame && ctx->coded_frame->pts != AV_NOPTS_VALUE)
pkt.pts = av_rescale_q(ctx->coded_frame->pts, ctx->time_base, vid_st->time_base);
if(pkt.size > VIDEO_BUF_LEN) {
avformat_free_context(oc);
oc = NULL;
return MRET_ERR_BUFSIZE;
}
if(ctx->coded_frame->key_frame)
pkt.flags |= AV_PKT_FLAG_KEY;
pkt.data = video_buf;
}
if(av_interleaved_write_frame(oc, &pkt) < 0) {
avformat_free_context(oc);
oc = NULL;
// yeah, err might not be a file error, but if it isn't, it's a
// coding error rather than a user-controllable error
// and better resolved using debugging
return MRET_ERR_FERR;
}
return MRET_OK;
}
#if LIBAVCODEC_VERSION_MAJOR > 56
/* FFmpeg depricated avcodec_encode_audio.
* It was removed completely in 3.0.
* This will at least get audio recording *working*
*/
static inline int MediaRecorderEncodeAudio(AVCodecContext *ctx,
AVPacket *pkt,
uint8_t *buf, int buf_size,
const short *samples)
{
AVFrame *frame;
av_init_packet(pkt);
int ret, samples_size, got_packet = 0;
pkt->data = buf;
pkt->size = buf_size;
if (samples) {
frame = frame = av_frame_alloc();
if (ctx->frame_size) {
frame->nb_samples = ctx->frame_size;
} else {
frame->nb_samples = (int64_t)buf_size * 8 /
(av_get_bits_per_sample(ctx->codec_id) *
ctx->channels);
}
frame->format = ctx->sample_fmt;
frame->channel_layout = ctx->channel_layout;
samples_size = av_samples_get_buffer_size(NULL, ctx->channels,
frame->nb_samples, ctx->sample_fmt, 1);
avcodec_fill_audio_frame(frame, ctx->channels, ctx->sample_fmt,
(const uint8_t *)samples, samples_size, 1);
//frame->pts = AV_NOPTS_VALUE;
} else {
frame = NULL;
}
ret = avcodec_encode_audio2(ctx, pkt, frame, &got_packet);
if (!ret && got_packet && ctx->coded_frame) {
ctx->coded_frame->pts = pkt->pts;
ctx->coded_frame->key_frame = !!(pkt->flags & AV_PKT_FLAG_KEY);
}
if (frame && frame->extended_data != frame->data)
av_freep(&frame->extended_data);
return ret;
}