September 2024 - ffmpeg-devel

[PATCH] avfilter/colorize: add speed option
by Yannis Gerlach 25 Jan '26

25 Jan '26

The speed option allows to have a constant (per frame) change of hue. This allows for an easy way of creating an color changing effect without relying on somewhat complicated expressions. Signed-off-by: Yannis Gerlach <ffmpeg(a)ygerlach.de> --- libavfilter/vf_colorize.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/libavfilter/vf_colorize.c b/libavfilter/vf_colorize.c index e6c563e3e2..ad8577c8fd 100644 --- a/libavfilter/vf_colorize.c +++ b/libavfilter/vf_colorize.c @@ -29,6 +29,7 @@ typedef struct ColorizeContext { float saturation; float lightness; float mix; + float speed; int depth; int c[3]; @@ -205,6 +206,13 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *frame) ff_filter_execute(ctx, do_slice, frame, NULL, FFMIN(s->planeheight[1], ff_filter_get_nb_threads(ctx))); + s->hue += s->speed; + if (s->hue < 0.f) { + s->hue += 360.f; + } else if(s->hue > 360.f) { + s->hue -= 360.f; + } + return ff_filter_frame(ctx->outputs[0], frame); } @@ -263,10 +271,11 @@ static const AVFilterPad colorize_inputs[] = { #define VF AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM static const AVOption colorize_options[] = { - { "hue", "set the hue", OFFSET(hue), AV_OPT_TYPE_FLOAT, {.dbl=0}, 0, 360, VF }, - { "saturation", "set the saturation", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl=0.5},0, 1, VF }, - { "lightness", "set the lightness", OFFSET(lightness), AV_OPT_TYPE_FLOAT, {.dbl=0.5},0, 1, VF }, - { "mix", "set the mix of source lightness", OFFSET(mix), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, VF }, + { "hue", "set the hue", OFFSET(hue), AV_OPT_TYPE_FLOAT, {.dbl=0}, 0, 360, VF }, + { "saturation", "set the saturation", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, VF }, + { "lightness", "set the lightness", OFFSET(lightness), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, VF }, + { "mix", "set the mix of source lightness", OFFSET(mix), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 1, VF }, + { "speed", "set the change of hue per frame", OFFSET(speed), AV_OPT_TYPE_FLOAT, {.dbl=0}, -180, 180, VF }, { NULL } }; -- 2.34.1

2 4

[PATCH 1/5] avcodec/get_buffer: Use av_buffer_mallocz() for audio same as its done for video
by Michael Niedermayer 03 Jul '25

03 Jul '25

Fixes: Use of uninintialized value Fixes: 70993/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AAC_LATM_fuzzer-6378949754552320 For the AAC/USAC/SBR code which reads uninitialized memory, it would be good, if it did not a fix for that is welcome! Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> --- libavcodec/get_buffer.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/libavcodec/get_buffer.c b/libavcodec/get_buffer.c index ff19f61e86b..7e705380fc6 100644 --- a/libavcodec/get_buffer.c +++ b/libavcodec/get_buffer.c @@ -142,7 +142,10 @@ static int update_frame_pool(AVCodecContext *avctx, AVFrame *frame) if (ret < 0) goto fail; - pool->pools[0] = av_buffer_pool_init(pool->linesize[0], NULL); + pool->pools[0] = av_buffer_pool_init(pool->linesize[0], + CONFIG_MEMORY_POISONING ? + NULL : + av_buffer_allocz); if (!pool->pools[0]) { ret = AVERROR(ENOMEM); goto fail; -- 2.46.0

3 13

[PATCH v2 1/3] avcodec/libaribb24: change new lines to \n in ASS header
by Kacper Michajłow 17 May '25

17 May '25

Fixes remaining \r\n is ASS header after 57c545090d. Signed-off-by: Kacper Michajłow <kasper93(a)gmail.com> --- libavcodec/libaribb24.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/libavcodec/libaribb24.c b/libavcodec/libaribb24.c index 009c995ded..0e4fa9935c 100644 --- a/libavcodec/libaribb24.c +++ b/libavcodec/libaribb24.c @@ -96,13 +96,13 @@ static int libaribb24_generate_ass_header(AVCodecContext *avctx) font_size = get_profile_font_size(avctx); avctx->subtitle_header = av_asprintf( - "[Script Info]\r\n" - "; Script generated by FFmpeg/Lavc%s\r\n" - "ScriptType: v4.00+\r\n" - "PlayResX: %d\r\n" - "PlayResY: %d\r\n" - "\r\n" - "[V4+ Styles]\r\n" + "[Script Info]\n" + "; Script generated by FFmpeg/Lavc%s\n" + "ScriptType: v4.00+\n" + "PlayResX: %d\n" + "PlayResY: %d\n" + "\n" + "[V4+ Styles]\n" /* ASSv4 header */ "Format: Name, " @@ -113,7 +113,7 @@ static int libaribb24_generate_ass_header(AVCodecContext *avctx) "Spacing, Angle, " "BorderStyle, Outline, Shadow, " "Alignment, MarginL, MarginR, MarginV, " - "Encoding\r\n" + "Encoding\n" "Style: " "Default," /* Name */ @@ -124,11 +124,11 @@ static int libaribb24_generate_ass_header(AVCodecContext *avctx) "0,0," /* Spacing, Angle */ "%d,1,0," /* BorderStyle, Outline, Shadow */ "%d,10,10,10," /* Alignment, Margin[LRV] */ - "0\r\n" /* Encoding */ + "0\n" /* Encoding */ - "\r\n" - "[Events]\r\n" - "Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text\r\n", + "\n" + "[Events]\n" + "Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text\n", !(avctx->flags & AV_CODEC_FLAG_BITEXACT) ? AV_STRINGIFY(LIBAVCODEC_VERSION) : "", plane_width, plane_height, ASS_DEFAULT_FONT, font_size, ASS_DEFAULT_COLOR, -- 2.43.0

4 5

[RFC/PATCH] MV-HEVC decoding
by Anton Khirnov 12 May '25

12 May '25

Hi, this patchset adds support for decoding MV-HEVC (multiview, typically used for stereoscopic 3D video). The code currently supports at most two views, with the second depending on the first (i.e. using it for inter prediction). Both views have to be of the same size; the spec allows them not to be, but I'm not supporting that, or a lot of other insanity in there. The frames for both views are interleaved in the output, with side data indicating which one has which view ID. If the relevant SEI is present, we also add side data indicating which view is left or right. Since the API caller needs to request the views it wants, and all the multilayer configuration is potentially per-sequence, there needs to be a negotiation process where the decoder exports available views and the caller picks from them. I'm implementing this by adding several decoder-private AVOptions, meant to be accessed from get_format(). Some of them are read-only and the caller reads available view IDs (and positions, if known) from them. The others are writable, and are used to receive the caller's choice. One potential point of discussion is that the caller may want to know the view ID in its get_buffer2() callback (this is actually used in the ffmpeg CLI code). One potential option is attaching the view ID side data before calling get_buffer2(), but that is quite unusual and would require changes to the progress frame API. For now, I'm using another read-only decoder-private AVOption for that. Opinions welcome. I'm also adding nontrivial support for this to ffmpeg CLI, which mainly means extending stream specifiers to allow specifying individual views, e.g. "-map 0:v:0:vpos:left" selects the left view (see TFM for more details). The set should be mostly functional, I'm tagging it as RFC because there's a little polishing left to do, and a few tests left to add. Testing and review welcome. You can fetch the set from git://git.khirnov.net/libav, branch 'mv-hevc'. Cheers, -- Anton Khirnov

6 75

[RFC] STF 2025
by Michael Niedermayer 25 Apr '25

25 Apr '25

Hi all Before this is forgotten again, better start some dicsussion too early than too late I propose that if we have the oppertunity again next year to receive a grant from STF. That we use it to fund: * Paul to work on FFmpeg full time. My idea here is that he can work on whatever he likes in FFmpeg (so its not full time employment for specific work but simply full time employment for him to work on whatever he likes in FFmpeg any way he likes) Paul is the 2nd largest contributor to FFmpeg (git shortlog -s -n) * Fund administrative / maintainance work (one example is the mailman upgrade that is needed with the next OS upgrade on one of our servers (this is not as trivial as one might expect). Another example here may be some git related tools if we find something that theres a broad consensus about. * Fund maintaince on the bug tracker, try to reproduce bugs, ask users to provide reproduceable cases, close bugs still unreproduceable, ... ATM we have over 2000 "new" bugs that are not even marked as open * Fund professional real live presence on multimedia / FOSS / buisness related events. we already refund individuals but i think we are lacking on the organizational side. We should also have on these events at least one person who can awnser developer/user questions and someone who can awnser buisness questions (on buisness related events). Also we need some eye catching things there, a big screen/projector that plays some real time filtered version from a camera. Or maybe have more people remotely be available from the FFmpeg team through real time streaming (as in, if someone wants to be on some event but cant physically go there, we could put a notebook on the table facing visitors showing something like a video chat. Also we need more cute girls on these events, everything i hear its 100% male geeks/hackers. Also a "24/7" realtime stream from any booth would be nice Thanks -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The real ebay dictionary, page 1 "Used only once" - "Some unspecified defect prevented a second use" "In good condition" - "Can be repaird by experienced expert" "As is" - "You wouldnt want it even if you were payed for it, if you knew ..."

12 67

[PATCH] avcodec/vp8: Maintain consistency of frame pointers
by Andreas Rheinhardt 21 Apr '25

21 Apr '25

Right now it is possible for the pointer for the current frame to be set in the context even when it could not be properly set up; this does not influence various the ordinary ref frames, but only VP8Context.prev_frame. And since this code has been ported to the ProgressFrame API in d48d7bc434f30dfbdf346f16715e4f2044b3e000, this leads to segfaults, because the ProgressFrame API is less forgiving than the ThreadFrame API (waiting on an uninitialized ProgressFrame segfaults, waiting on an uninitialized ThreadFrame is a no-op (the code behaves as if frame-threading is not in use)). Fix this by maintaining the consistency of the frame pointers in the context (by setting them later). Fixes: NULL pointer dereference Fixes: 68192/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VP8_fuzzer-6180311026171904 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)outlook.com> --- libavcodec/vp8.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c index d6df018655..8945447eb6 100644 --- a/libavcodec/vp8.c +++ b/libavcodec/vp8.c @@ -541,9 +541,12 @@ static int vp7_fade_frame(VP8Context *s, int alpha, int beta) /* preserve the golden frame, write a new previous frame */ if (s->framep[VP8_FRAME_GOLDEN] == s->framep[VP8_FRAME_PREVIOUS]) { - s->framep[VP8_FRAME_PREVIOUS] = vp8_find_free_buffer(s); - if ((ret = vp8_alloc_frame(s, s->framep[VP8_FRAME_PREVIOUS], 1)) < 0) + VP8Frame *prev_frame = vp8_find_free_buffer(s); + + ret = vp8_alloc_frame(s, prev_frame, 1); + if (ret < 0) return ret; + s->framep[VP8_FRAME_PREVIOUS] = prev_frame; dst = s->framep[VP8_FRAME_PREVIOUS]->tf.f; @@ -2699,7 +2702,7 @@ int vp78_decode_frame(AVCodecContext *avctx, AVFrame *rframe, int *got_frame, &s->frames[i] != s->framep[VP8_FRAME_ALTREF]) vp8_release_frame(&s->frames[i]); - curframe = s->framep[VP8_FRAME_CURRENT] = vp8_find_free_buffer(s); + curframe = vp8_find_free_buffer(s); if (!s->colorspace) avctx->colorspace = AVCOL_SPC_BT470BG; @@ -2723,6 +2726,7 @@ int vp78_decode_frame(AVCodecContext *avctx, AVFrame *rframe, int *got_frame, if ((ret = vp8_alloc_frame(s, curframe, referenced)) < 0) goto err; + s->framep[VP8_FRAME_CURRENT] = curframe; if (s->keyframe) curframe->tf.f->flags |= AV_FRAME_FLAG_KEY; else -- 2.40.1

1 1

[PATCH] avcodec/ffv1enc: Use version 3 by default (CRCs by default)
by Michael Niedermayer 12 Apr '25

12 Apr '25

Version 3 is since 2013 (FFmpeg 2.1) non experimental so should be widely supported Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> --- libavcodec/ffv1enc.c | 3 ++- tests/fate/vcodec.mak | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c index c521b7d445..ce66d0e4d6 100644 --- a/libavcodec/ffv1enc.c +++ b/libavcodec/ffv1enc.c @@ -542,7 +542,8 @@ static av_cold int encode_init(AVCodecContext *avctx) return AVERROR(EINVAL); } s->version = avctx->level; - } + } else if (s->version < 3) + s->version = 3; if (s->ec < 0) { s->ec = (s->version >= 3); diff --git a/tests/fate/vcodec.mak b/tests/fate/vcodec.mak index 452246689e..91b46034c8 100644 --- a/tests/fate/vcodec.mak +++ b/tests/fate/vcodec.mak @@ -168,7 +168,7 @@ FATE_VCODEC-$(call ENCDEC, FFV1, AVI) += ffv1 ffv1-v0 \ ffv1-v3-yuv420p ffv1-v3-yuv422p10 ffv1-v3-yuv444p16 \ ffv1-v3-bgr0 ffv1-v3-rgb48 fate-vsynth%-ffv1: ENCOPTS = -slices 4 -fate-vsynth%-ffv1-v0: CODEC = ffv1 +fate-vsynth%-ffv1-v0: ENCOPTS = -level 0 fate-vsynth%-ffv1-v3-yuv420p: ENCOPTS = -level 3 -pix_fmt yuv420p fate-vsynth%-ffv1-v3-yuv422p10: ENCOPTS = -level 3 -pix_fmt yuv422p10 \ -sws_flags neighbor+bitexact -- 2.23.0

3 4

[PATCH 1/3] lavu/dict: add AV_DICT_DEDUP
by rcombs 06 Apr '25

06 Apr '25

This is useful when multiple metadata inputs may set the same value (e.g. both a container-specific header and an ID3 tag). --- libavutil/dict.c | 11 +++++++++++ libavutil/dict.h | 1 + 2 files changed, 12 insertions(+) diff --git a/libavutil/dict.c b/libavutil/dict.c index 9d3d96c58b..aa98f753a3 100644 --- a/libavutil/dict.c +++ b/libavutil/dict.c @@ -76,6 +76,17 @@ int av_dict_set(AVDictionary **pm, const char *key, const char *value, if (!(flags & AV_DICT_MULTIKEY)) { tag = av_dict_get(m, key, NULL, flags); + } else if (flags & AV_DICT_DEDUP) { + while ((tag = av_dict_get(m, key, tag, flags))) { + if ((!value && !tag->value) || + (value && tag->value && !strcmp(value, tag->value))) { + if (flags & AV_DICT_DONT_STRDUP_KEY) + av_free((void*)key); + if (flags & AV_DICT_DONT_STRDUP_VAL) + av_free((void*)value); + return 0; + } + } } if (flags & AV_DICT_DONT_STRDUP_KEY) copy_key = (void *)key; diff --git a/libavutil/dict.h b/libavutil/dict.h index 0d1afc6c64..a957b48bf3 100644 --- a/libavutil/dict.h +++ b/libavutil/dict.h @@ -75,6 +75,7 @@ #define AV_DICT_APPEND 32 /**< If the entry already exists, append to it. Note that no delimiter is added, the strings are simply concatenated. */ #define AV_DICT_MULTIKEY 64 /**< Allow to store several equal keys in the dictionary */ +#define AV_DICT_DEDUP 128 /**< If inserting a value that already exists for a key, do nothing. Only relevant with AV_DICT_MULTIKEY. */ typedef struct AVDictionaryEntry { char *key; -- 2.37.1

2 3

[PATCH] avformat/dashenc: add option to set live DASH stream start time
by Jerome Berclaz 01 Apr '25

01 Apr '25

When encoding a DASH stream in live ("dynamic") mode without timeline, the field "availabilityStartTime" is set to the system time. The new options allows the libavformat user to pass a specific time to be used instead of the system time. --- libavformat/dashenc.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/libavformat/dashenc.c b/libavformat/dashenc.c index d4a6fe0304..e586e934cb 100644 --- a/libavformat/dashenc.c +++ b/libavformat/dashenc.c @@ -204,6 +204,7 @@ typedef struct DASHContext { AVRational min_playback_rate; AVRational max_playback_rate; int64_t update_period; + int64_t start_time_ms; } DASHContext; static const struct codec_string { @@ -2108,7 +2109,12 @@ static int dash_write_packet(AVFormatContext *s, AVPacket *pkt) os->last_pts = pkt->pts; if (!c->availability_start_time[0]) { - int64_t start_time_us = av_gettime(); + int64_t start_time_us; + if (c->start_time_ms) { + start_time_us = c->start_time_ms * 1000; + } else { + start_time_us = av_gettime(); + } c->start_time_s = start_time_us / 1000000; format_date(c->availability_start_time, sizeof(c->availability_start_time), start_time_us); @@ -2393,6 +2399,7 @@ static const AVOption options[] = { { "seg_duration", "segment duration (in seconds, fractional value can be set)", OFFSET(seg_duration), AV_OPT_TYPE_DURATION, { .i64 = 5000000 }, 0, INT_MAX, E }, { "single_file", "Store all segments in one file, accessed using byte ranges", OFFSET(single_file), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, E }, { "single_file_name", "DASH-templated name to be used for baseURL. Implies storing all segments in one file, accessed using byte ranges", OFFSET(single_file_name), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, E }, + { "start_time_ms", "Stream start time in epoch milliseconds", OFFSET(start_time_ms), AV_OPT_TYPE_INT64, {.i64 = 0}, 0, UINT64_MAX, E, .unit = "ms"}, { "streaming", "Enable/Disable streaming mode of output. Each frame will be moof fragment", OFFSET(streaming), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, E }, { "target_latency", "Set desired target latency for Low-latency dash", OFFSET(target_latency), AV_OPT_TYPE_DURATION, { .i64 = 0 }, 0, INT_MAX, E }, { "timeout", "set timeout for socket I/O operations", OFFSET(timeout), AV_OPT_TYPE_DURATION, { .i64 = -1 }, -1, INT_MAX, .flags = E }, -- 2.43.0

1 1

[PATCH v2 0/6] WebRTC sub-second live streaming support
by Michael Riedl 23 Mar '25

23 Mar '25

This patch series adds support for WHIP and WHEP (WebRTC-HTTP ingestion protocol and WebRTC-HTTP egress protocol). It supersedes the previous patch series. Thank you all for your feedback and suggestions! The WHIP and WHEP are defined in the following draft RFCs: - WHIP: https://datatracker.ietf.org/doc/draft-ietf-wish-whip - WHEP: https://datatracker.ietf.org/doc/draft-murillo-whep/ The implementation builds on FFmpeg's existing RTP, HTTP and SDP support as well as libdatachannel for the rest. This library is much more lightweight than other libraries (e.g. libwebrtc). At the same time, using this library avoids reimplementing parts of WebRTC in FFmpeg. This patch series was tested with WebRTC servers Dolby.io (formerly Millicast) and SRS (https://github.com/ossrs/srs) Using a local server, an end-to-end latency of 3 frames (50 ms at 60 fps) was measured with video-only output, and 12 frames (200 ms at 60 fps) with audio and video output. Using a DeckLink input device, an end-to-end latency of about 4 frames (70 ms at 60 fps) was measured with both, video-only and audio and video output. Using a remote server, only the RTT is added to the end-to-end latency. For example, using a server in Amsterdam (The Netherlands) from a location in Salzburg (Austria), with RTT=18ms, an end-to-end latency of 50+18=68 ms was measured for video-only output. Michael Riedl (6): libavformat/http: expose actual Location header value libavformat/sdp: remove whitespaces in fmtp configure: add libdatachannel as external library libavformat/webrtc: add common code for WebRTC streaming libavformat/webrtc_demux: add WebRTC-HTTP egress protocol (WHEP) demuxer libavformat/webrtc_mux: add WebRTC-HTTP ingestion protocol (WHIP) muxer Changelog | 5 + MAINTAINERS | 1 + configure | 7 + doc/demuxers.texi | 22 ++ doc/muxers.texi | 21 ++ libavformat/Makefile | 2 + libavformat/allformats.c | 2 + libavformat/http.c | 1 + libavformat/sdp.c | 4 +- libavformat/webrtc.c (new) | 410 +++++++++++++++++++++++++++++++ libavformat/webrtc.h (new) | 70 ++++++ libavformat/webrtc_demux.c (new) | 246 +++++++++++++++++++ libavformat/webrtc_mux.c (new) | 273 ++++++++++++++++++++ 13 files changed, 1062 insertions(+), 2 deletions(-) create mode 100644 libavformat/webrtc.c create mode 100644 libavformat/webrtc.h create mode 100644 libavformat/webrtc_demux.c create mode 100644 libavformat/webrtc_mux.c -- 2.39.2

5 7