To add, update or remove a test file, please update the list below.
Please provide full reference and steps to generate the test file so that any people can regenerate or update the file in the future.
WebM encode of bear.1280x720.mp4 resized to 320x240.
The video track of bear-320x240.webm.
The audio track of bear-320x240.webm.
VP9 video only WebM file.
VP9 Video with Opus Audio.
Opus Audio only WebM file.
WebM VP8 video with WebVTT subtitle track.
AAC audio and H264 high profile video.
Fragmented bear_1280x720.mp4 with text track containing srt from bear-vp8-webvtt.webm as a ‘subt’ handler type.
Just the first initialization segment of bear-1280x720_av_frag.mp4, modified to have the mvhd version 0 32-bit duration field set to all 1's.
A truncated audio/video file with audio packet timestamps of -1. We need to ensure that these packets aren't dropped.
Fragmented mp4 of noise encoded with xHE-AAC, from xHE-AAC samples in Android CTS, using ffmpeg version 4.2.2 (where nofillin lets audio nonkeyframes in input be indicated the same in output, unlike more recent tip-of-tree ffmpeg's operation with this option) to remux, unfortunately with empty MOOV not giving real duration:
ffmpeg -fflags nofillin -i noise_2ch_48khz_aot42_19_lufs_mp4.m4a -acodec copy -t 1 -movflags frag_keyframe+empty_moov+default_base_moof noise-xhe-aac.mp4
Unfragmented audio-only 44.1kHz FLAC in MP4 file, created using:
ffmpeg -i bear-1280x720.mp4 -map 0:0 -acodec flac -strict -2 bear-flac.mp4
Note: “-strict -2” was required because current ffmpeg libavformat version 57.75.100 indicates that flac in MP4 support is experimental.
Fragmented audio-only 44.1kHz FLAC in MP4 file, created using:
ffmpeg -i bear-flac.mp4 -acodec copy -strict -2 -movflags frag_keyframe+empty_moov+default_base_moof bear-flac_frag.mp4
Unfragmented audio-only high-sample-rate FLAC in MP4 file, created using:
ffmpeg -i bear-1280x720.mp4 -map 0:0 -acodec flac -strict -2 -ar 192000 bear-flac-192kHz.mp4
Fragmented audio-only high-sample-rate FLAC in MP4 file, created using:
ffmpeg -i bear-flac-192kHz.mp4 -acodec copy -strict -2 -movflags frag_keyframe+empty_moov+default_base_moof bear-flac-192kHz_frag.mp4
Unfragmented audio-only 44.1kHz FLAC in MP4 file, created using:
ffmpeg -i sfx.flac -map 0:0 -acodec copy -strict -2 sfx-flac.mp4
Fragmented audio-only 44.1kHz FLAC in MP4 file, created using:
ffmpeg -i sfx.flac -map 0:0 -acodec copy -strict -2 -movflags frag_keyframe+empty_moov+default_base_moof sfx-flac_frag.mp4
Unless noted otherwise, the codec string is av01.0.04M.08 for 8-bit files, and av01.0.04M.10 for 10-bit files.
Not an AV1 file, but all of the following commands rely on this file. It was created using vpxdec with the following command:
vpxdec path/to/chrome/src/media/test/data/bear-vp9.webm -o bear.y4m
Created using FFmpeg with the following commands:
ffmpeg -i bear.y4m -vcodec libaom-av1 -strict -2 -y -f mp4 -b:v 50k \ bear-av1-slowstart.mp4 ffmpeg -i bear-av1-slowstart.mp4 -vcodec copy -strict -2 -y -f mp4 \ -movflags frag_keyframe+empty_moov+default_base_moof+faststart bear-av1.mp4
Similar to the above but using aomenc for encoding with the --monochrome option.
Created using aomenc with the following command:
aomenc bear.y4m --lag-in-frames=0 --target-bitrate=50 --fps=30000/1001 \ --cpu-used=8 --test-decode=fatal -o bear-av1.webm
Created using FFmpeg and aomenc with the following commands:
ffmpeg -i bear.y4m -vf scale=-1:360 -f rawvideo bear_360P.yuv aomenc bear_360P.yuv -w 480 -h 360 --fps=30000/1001 --cpu-used=8 \ --lag-in-frames=0 --test-decode=fatal --target-bitrate=50 \ -o bear-av1-480x360.webm
Created using FFmpeg and aomenc with the following commands:
ffmpeg -i bear.y4m -vf scale=-1:480 -f rawvideo bear_480P.yuv aomenc bear_480P.yuv -w 640 -h 480 --fps=30000/1001 --cpu-used=8 \ --lag-in-frames=0 --test-decode=fatal --target-bitrate=50 \ -o bear-av1-640x480.webm
Created using FFmpeg with the following command:
ffmpeg -i bear-av1-320x180-10bit.webm -vcodec copy -f mp4 \ -movflags frag_keyframe+empty_moov+default_base_moof+faststart \ bear-av1-320x180-10bit.mp4
Created using vpxdec and aomenc with the following commands:
vpxdec bear-320x180-hi10p-vp9.webm -o bear-320x180-10bit.y4m aomenc bear-320x180-10bit.y4m --lag-in-frames=0 --target-bitrate=50 \ --fps=30000/1001 --cpu-used=8 --bit-depth=10 --test-decode=fatal \ -o bear-av1-320x180-10bit.webm
Created by combining bear-av1.mp4 and bear-opus.mp4.
ffmpeg -i bear-av1.mp4 -i bear-opus.mp4 -c copy -strict -2 \ -movflags frag_keyframe+empty_moov+default_base_moof \ bear-av1-opus.mp4
Note: “-strict -2” was required because the current ffmpeg version has support for OPUS in MP4 as experimental.
AV1 data that has spatial and temporal layers. This is the same as av1-1-b8-22-svc-L2T2.ivf in libaom test vectors. The video license is libaom LICENSE.
AV1 data that contains frames with show_existing_frame=1. This is the same as 00000592.ivf in https://people.xiph.org/~tterribe/av1/samples-all/. The video license is libaom LICENSE.
The first frame of blackwhite_yuv444p.mp4 coded in AV1 by the following command. ffmpeg -i blackwhite_yuv444p.mp4 -strict -2 -vcodec av1 -vframes 1 blackwhite_yuv444p-frame.av1.ivf
AV1 data where film grain feature is used. This is the same as av1-1-b8-23-film_grain-50.ivf in libaom test vectors. The video license is libaom LICENSE.
WebM VP8 video with alpha channel.
WebM VP8 video with alpha channel and odd dimensions.
The first I frame of a 160x240 re-encode of bear-320x240.webm.
The first I frame of a 320x120 re-encode of bear-320x240.webm.
The first I frame of bear-320x240.webm.
The second P frame of bear-320x240.webm.
The first I frame of a 320x480 re-encode of bear-320x240.webm.
The first I frame of a 640x240 re-encode of bear-320x240.webm.
A copy of vp8-I-frame-320x240 w/ all bytes XORed w/ 0xA5.
Header, Info, & Tracks element from bear-320x240.webm slightly corrupted so it looks like there are no tracks.
Has the same headers as bear-320x240.webm but the first cluster of this file is the second cluster of bear-320x240.webm. This creates the situation where the media data doesn't start at time 0.
bear-320x240.webm's initialization segment followed by “CORRUPTED\n”
bear-320x240.webm remuxed w/o a duration and using clusters with unknown sizes.
ffmpeg -i bear-320x240.webm -acodec copy -vcodec copy -f webm pipe:1 > bear-320x240-live.webm
a WebM file containing color metadata in MKV/WebM Colour element copied from libwebm/testing/testdata/colour.webm
A 960x540 H.264 mp4 video with 4 color blocks (Y,R,G,B) in every frame. The video playback looks like a still image. An image of 4 color blocks (.png file) is first created by Windows Paint.exe. This image is then used as a basic video frame in making this 2-second video from Mac iStopMotion.
Actual video frames are the same as four-colors.mp4, except it specifies an aspect of 4x3 in mp4 meta data.
Actual video frames are the same as four-colors.mp4, except it specifies a rotation of 90 degrees in mp4 meta data.
Actual video frames are the same as four-colors.mp4, except it specifies a rotation of 180 degrees in mp4 meta data.
Actual video frames are the same as four-colors.mp4, except it specifies a rotation of 270 degrees in mp4 meta data.
A 962x540 raw YUV single frame video with 4 color blocks (Y,R,G,B) and a GL incompatible stride. Converted from four-colors.mp4 using ffmpeg:
ffmpeg -i four-colors.mp4 -vf "scale=w=962:h=540,format=yuv420p" -frames:v 1 four-colors-incompatible-stride.y4m
A 960x540 vp9 video with 4 color blocks (Y,R,G,B) in every frame. This is converted from four-colors.mp4 by ffmpeg.
A 960x540 yuva420p vp9 video with 4 color blocks (Y,R,G,B) in every frame. This is converted from four-colors.mp4 by adding an opacity of 0.5 using ffmpeg.
VP9 encoded video with profile 2 (10-bit, 4:2:0). Codec string: vp09.02.10.10.01.02.02.02.00.
Init segment for a VP9.2 HDR in MP4 file; from https://crbug.com/1102200#c6. The SmDm and CoLL boxes have been added using mp4edit:
mp4edit.exe --insert moov/trak/mdia/minf/stbl/stsd/vp09:smdm.bin
--insert moov/trak/mdia/minf/stbl/stsd/vp09:coll.bin
vp9-hdr-init-segment.mp4 fixed.mp4
smdm.bin and coll.bin generated with program from https://crbug.com/1123430#c5.
The first four frames of the H.264 version of bear-av1-320x180-10bit.mp4 created using the following command. ffmpeg -i bear-av1-320x180-10bit.mp4 -vcodec h264 -vframes 4 bear-320x180-10bit-4frames.h264 The file is then split into bitstreams each of which contains a single frame, so that they contain frames as below. bear-320x180-10bit-frame-0.h264: SPS+PPS+Single IDR bear-320x180-10bit-frame-1.h264: B bear-320x180-10bit-frame-2.h264: B bear-320x180-10bit-frame-3.h264: P
Duration of each packet is (1024/44100 Hz), approximately 23.22 ms.
Just first 2 video frames of bear-640x360-v_frag.mp4, created with:
ffmpeg -i bear-640x360-v_frag.mp4 -vcodec copy -movflags frag_keyframe+empty_moov+default_base_moof \
-vframes 2 bear-640x360-v-2frames_frag.mp4
It‘s 1 keyframe + 1 non-keyframe, with container’s frame keyframe-ness correct.
This is bear-640x360-v-2frames_frag.mp4, with manually updated trun.first_sample_flags: s/0x02000000/0x01010000 (first frame is non-sync-sample, depends on another frame, mismatches compressed h264 first frame's keyframe-ness).
This is bear-640x360-v-2frames_frag.mp4, with manually updated tfhd.default_sample_flags: s/0x01010000/0x02000000 (second frame is sync-sample, doesn‘t depend on other frames, mismatches compressed h264 second frame’s nonkeyframe-ness).
[1] Test key ID: 30313233343536373839303132333435
[2] Test key: ebdd62f16814d27b68ef122afce4ae3c
[3] KeyIds and Keys are created by left rotating key ID [1] and key [2] using std::rotate for every new crypto period. This is only for testing. The actual key rotation algorithm is often much more complicated.
A fragmented MP4 version of the audio track of bear-1280x720.mp4 encrypted (ISO CENC) using key ID [1] and key [2].
A fragmented MP4 version of the audio track of bear-1280x720.mp4 encrypted (ISO CENC) using key ID [1] and key [2] with key rotation [3].
Same as bear-1280x720-a_frag-cenc.mp4 but no fragments are encrypted.
A fragmented MP4 version of the video track of bear-1280x720.mp4 encrypted (ISO CENC) using key ID [1] and key [2].
A fragmented MP4 version of the video track of bear-1280x720.mp4 encrypted (ISO CENC) using key ID [1] and key [2] with key rotation [3].
Same as bear-1280x720-v_frag-cenc.mp4 but no fragments are encrypted.
An invalid file similar to bear-1280x720-a_frag-cenc.mp4 but has no saiz and saio boxes. To save space, it has only one encrypted sample.
Bear video with VP9 codec in MP4 container. Generated with Shaka Packager at 1e2da22c8809c17cc4dfdb45924ec45e42058393:
packager in=bear-vp9.webm,stream=video,out=bear-320x240-v_frag-vp9.mp4
Same as above, with video encrypted using key ID [1] and key [2]. Generated with Shaka Packager at 1e2da22c8809c17cc4dfdb45924ec45e42058393:
packager in=bear-vp9.webm,stream=video,out=bear-320x240-v_frag-vp9-cenc.mp4
--enable_fixed_key_encryption --key_id 30313233343536373839303132333435
--key ebdd62f16814d27b68ef122afce4ae3c --clear_lead 0
--pssh 0000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000000000467073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000261210303132333435363738393031323334351A00221030313233343536373839303132333435
bear-320x240-16x9-aspect.webm with audio & video encrypted using key ID [1] and key [2]
bear-320x240.webm with audio & video encrypted using key ID [1] and key [2].
Same as bear-320x240-av_enc-av.webm but with no frames in the first second encrypted.
Same as bear-320x240-av_enc-av.webm but with no frames encrypted.
Encrypted Bear video with VP9 codec (profile 2) in WebM container, using key ID [1] and key [2]. Codec string: vp09.02.10.10.01.02.02.02.00. Generated with Shaka Packager at 4ba5bec66054cfd4af13c07ac62a97f1a1a2e5f9:
packager in=bear-320x240-vp9_profile2.webm,stream=video,out=bear-320x240-v-vp9_profile2_subsample_cenc-v.webm
--enable_fixed_key_encryption --key_id 30313233343536373839303132333435
--key ebdd62f16814d27b68ef122afce4ae3c --clear_lead 0
--pssh 0000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000000000467073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000261210303132333435363738393031323334351A00221030313233343536373839303132333435
Same as above, in MP4 container. Codec string: vp09.02.10.10.01.02.02.02.00. Generated with Shaka Packager at 4ba5bec66054cfd4af13c07ac62a97f1a1a2e5f9:
packager in=bear-320x240-vp9_profile2.webm,stream=video,out=bear-320x240-v-vp9_profile2_subsample_cenc-v.mp4
--enable_fixed_key_encryption --key_id 30313233343536373839303132333435
--key ebdd62f16814d27b68ef122afce4ae3c --clear_lead 0
--pssh 0000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000000000467073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000261210303132333435363738393031323334351A00221030313233343536373839303132333435
bear-640x360.webm with audio & video encrypted using key ID [1] and key [2].
bear-320x240.webm with video track encrypted using key ID [1] and key [2].
bear-320x240.webm with audio track encrypted using key ID [1] and key [2].
bear-320x240-video-only.webm encrypted using key ID [1] and key [2].
bear-vp9.webm VP9 video only encrypted using key ID [1] and key [2] with full sample encryption.
bear-vp9.webm VP9 video only encrypted using key ID [1] and key [2] with subsample encryption.
bear-opus.webm encrypted using key ID [1] and key [2].
bear-vp9-opus.webm with audio & video encrypted using key ID [1] and key [2].
bear-vp9-opus.webm with video track encrypted using key ID [1] and key [2].
A fragmented MP4 version of the audio track of bear-640x360.mp4 encrypted (ISO CENC) using key ID [1] and key [2].
Note: bear-640x360.mp4 file does not exist any more. Files encrypted from it has AAC audio and H264 high profile video (if applicable).
A fragmented MP4 version of the audio track of bear-640x360.mp4 encrypted (ISO CENC) using key ID [1] and key [2] with key rotation [3].
A fragmented MP4 version of the video track of bear-640x360.mp4 encrypted (ISO CENC) using key ID [1] and key [2] and with sample encryption auxiliary information in the beginning of mdat box.
Same as above, but with sample encryption information stored in SampleEncryption (‘senc’) box.
Same as above, but without saiz and saio boxes.
A fragmented MP4 version of the video track of bear-640x360.mp4 encrypted (ISO CENC) using key ID [1] and key [2] with key rotation [3].
Encrypted version of bear-flac.mp4, encrypted by Shaka Packager v2.1.0 using key ID [1] and key [2]. Sample encryption information stored in SampleEncryption (‘senc’) box (in each encrypted fragment).
packager in=bear-flac.mp4,stream=audio,output=bear-flac-cenc.mp4
--enable_raw_key_encryption
--clear_lead 0.5
--segment_duration 0.5
--keys label=:key_id=30313233343536373839303132333435:key=ebdd62f16814d27b68ef122afce4ae3c
--pssh 000000327073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000121210303132333435363738393031323334350000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000
Encrypted version of bear-opus.mp4, encrypted by Shaka Packager v2.3.0 using key ID [1] and key [2].
packager in=bear-opus.mp4,stream=audio,output=bear-opus-cenc.mp4
--enable_raw_key_encryption
--protection_scheme cenc
--clear_lead 0
--segment_duration 0.5
--keys label=:key_id=30313233343536373839303132333435:key=ebdd62f16814d27b68ef122afce4ae3c
--pssh 000000327073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000121210303132333435363738393031323334350000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000
bear-320x240-audio-only.webm encrypted using key ID [1] and key [2].
third_party/WebKit/LayoutTests/media/resources/frame_size_change.webm encrypted using key ID [1] and key [2].
Unless noted otherwise, the codec string is av01.0.04M.08 for 8-bit files, and av01.0.04M.10 for 10-bit files.
vpxdec media/test/data/bear-vp9.webm -o bear.y4m aomenc -o bear.ivf -p 2 --target-bitrate=150 bear.y4m --limit=1 --ivf tail -c +45 bear.ivf > av1-I-frame-320x240
Same as av1-I-frame-320x240 but using bear-1280x720.webm as input.
Same as av1-I-frame-320x240 with --monochrome and -b=[8,10,12] aomenc options.
Encrypted version of bear-av1.mp4. Encrypted by Shaka Packager built locally at commit 53aa775ea488c0ffd3a2e1cb78ad000154e414e1 using key ID [1] and key [2].
packager in=bear-av1.mp4,stream=video,output=bear-av1-cenc.mp4
--enable_raw_key_encryption --protection_scheme cenc --clear_lead 0
--keys label=:key_id=30313233343536373839303132333435:key=ebdd62f16814d27b68ef122afce4ae3c
--pssh 000000327073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000121210303132333435363738393031323334350000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000
Same as bear-av1-cenc.mp4, except that the output name is bear-av1-cenc.webm.
Same as bear-av1-cenc.mp4, except that the input name is bear-av1-320x180-10bit.mp4, and the output name is bear-av1-320x180-10bit-cenc.mp4.
Same as bear-av1-cenc.mp4, except that the input name is bear-av1-320x180-10bit.webm, and the output name is bear-av1-320x180-10bit-cenc.webm.
Encrypted versions of bear-640x360-v_frag.mp4, encrypted by Shaka Packager v2.0.0 using key ID [1] and key [2]. Sample encryption information stored in SampleEncryption (‘senc’) box (in each encrypted fragment).
Command: (replace both places of ‘cenc’ with desired scheme)
packager in=bear-640x360-v_frag.mp4,stream=video,output=bear-640x360-v-cenc.mp4
--enable_raw_key_encryption
--protection_scheme cenc
--clear_lead 0.5
--segment_duration 0.5
--keys label=:key_id=30313233343536373839303132333435:key=ebdd62f16814d27b68ef122afce4ae3c
--pssh 000000327073736800000000EDEF8BA979D64ACEA3C827DCD51D21ED000000121210303132333435363738393031323334350000003470737368010000001077EFECC0B24D02ACE33C1E52E2FB4B000000013031323334353637383930313233343500000000
The ‘pssh’ data includes entries for both Widevine and Common SystemID. It was generated from concatenating the output of:
shaka/packager/tools/pssh/pssh-box.py --widevine-system-id --key-id 30313233343536373839303132333435 --hex shaka/packager/tools/pssh/pssh-box.py --common-system-id --key-id 30313233343536373839303132333435 --hex
Produced using Apple's mediafilesegmenter tool with bear-1280x720.ts as input, with no encryption.
mediafilesegmenter -t 10 -start-segments-with-iframe -f 'output_clear/' bear-1280x720.ts
Produced using Apple's mediafilesegmenter tool also with bear-1280x720.ts as input, but with SAMPLE_AES encryption. (Additional TS packets were constructed manually and prepended to convey the encryption metadata, notably key id and IV).
mediafilesegmenter -S -P -k 'key_iv.bin' -t 10 -start-segments-with-iframe -f 'output/' bear-1280x720.ts
Same as bear-1280x720-hls.ts but with an extra TS packet prepended. This is the same as the first of the metadata packets in bear-1280x720-hls-sample-aes.ts. Its presence indicates that SAMPLE_AES encryption may occur later in the stream, and causes the initial audio and video configs to have an encryption_scheme (of AES-CBC). The actual data is unencrypted, which is indicated by the lack of key ID and IV. This ends up very similar to how clear leader of an otherwise encrypted stream can occur in MP4.
Additional containers derived from bear.ogv:
avconv -i bear.ogv -f ac3 -b 192k bear.ac3.avconv -i bear.ogv -f adts -strict experimental bear.adts.avconv -i bear.ogv -f aiff bear.aiff.avconv -i bear.ogv -f asf bear.asf.avconv -i bear.ogv -f avi -b 192k bear.avi.avconv -i bear.ogv -f eac3 bear.eac3.avconv -i bear.ogv -f flac bear.flac.avconv -i bear.ogv -f flv bear.flv.avconv -i bear.ogv -f h261 -s:0 cif bear.h261.avconv -i bear.ogv -f h263 -s:0 cif bear.h263.avconv -i bear.ogv -f mpegts bear.m2ts.avconv -i bear.ogv -f mjpeg bear.mjpeg.avconv -i bear.ogv -f mpeg bear.mpeg.avconv -i bear.ogv -f rm -b 192k bear.rm.avconv -i bear.ogv -f swf -an bear.swf.Using ffmpeg SVN-r0.5.9-4:0.5.9-0ubuntu0.10.04.1 @ WebKit r122718, generated with:
ffmpeg -i third_party/WebKit/LayoutTests/media/content/test-25fps.mp4 \
-vcodec copy -vbsf h264_mp4toannexb -an test-25fps.h264
JSON file that contains all metadata related to test-25fps.h264, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
ffmpeg git-2012-07-19-a8d8e86, libvpx ToT 7/19, chromium r147247, mkvextract v5.0.1
ffmpeg -i test-25fps.h264 -vcodec libvpx -an test-25fps.webm && \
mkvextract tracks test-25fps.webm 1:test-25fps.vp8 && rm test-25fps.webm
JSON file that contains all metadata related to test-25fps.vp8, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
avconv 9.16-6:9.16-0ubuntu0.14.04.1, vpxenc v1.3.0
avconv -i test-25fps.h264 -c:v rawvideo -pix_fmt yuv420p test-25fps_i420.yuv
vpxenc test-25fps_i420.yuv -o test-25fps.vp9 --codec=vp9 -w 320 -h 240 --ivf \
--profile=0 --kf-min-dist=0 --kf-max-dist=150 --lag-in-frames=24 \
--drop-frame=0 --target-bitrate=140 --cq-level=23 --min-q=4 --max-q=56 \
--static-thresh=1000 --arnr-maxframes=7 --arnr-strength=5 --arnr-type=3 \
--cpu-used=1 --good --tile-columns=1 --passes=2 --threads=1 --fps=25/1 \
--end-usage=cq --auto-alt-ref=1 --bias-pct=50 --minsection-pct=0 \
--maxsection-pct=2000 --undershoot-pct=100
JSON file that contains all metadata related to test-25fps.vp9, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
Similar to test-25fps.vp9, substituting the option --profile=0 with --profile=2 --bit-depth=10 to vpxenc. (Note that vpxenc must have been configured with the option --enable-vp9-highbitdepth).
JSON file that contains all metadata related to test-25fps.vp9_2, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
The av1 video whose content is the same as test-25fps.h264.
ffmpeg -i test-25fps.h264 -vcodec libaom-av1 test-25fps.av1.ivf
JSON file that contains all metadata related to test-25fps.av1.ivf, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
H.265/HEVC video whose content is the same as test-25fps.h264.
ffmpeg -i test-25fps.h264 -vcodec hevc test25fps.hevc
JSON file that contains all metadata related to test-25fps.hevc, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
10-bit H.265/HEVC video whose content is the same as test-25fps.h264 but converted to 10bpp.
ffmpeg -i test-25fps.h264 -vcodec hevc -pix_fmt yuv420p10le test25fps.hevc10
JSON file that contains all metadata related to test-25fps.hevc10, used by the video_decode_accelerator_tests. This includes the video codec, resolution and md5 checksums of individual video frames when converted to the I420 format.
1280x720 version of Big Buck Bunny https://peach.blender.org/ muxed with raw vp9 frames (versus superframes).
VP9 video with show_existing_frame flag. The original test stream comes from Android CTS.
ffmpeg -i vp90_2_17_show_existing_frame.vp9 -vcodec copy -an -f ivf \
vp90_2_17_show_existing_frame.vp9.ivf
Using ffmpeg version 0.8.6-4:0.8.6-0ubuntu0.12.04.1, generated with bear.mp4 (https://chromiumcodereview.appspot.com/10805089):
ffmpeg -i bear.mp4 -vcodec copy -vbsf h264_mp4toannexb -an bear.h264
Using ffmpeg version 0.8.6-4:0.8.6-0ubuntu0.12.04.1, generated with npot-video.mp4 (https://codereview.chromium.org/8342021):
ffmpeg -i npot-video.mp4 -vcodec copy -vbsf h264_mp4toannexb -an npot-video.h264
Using ffmpeg version 0.8.6-4:0.8.6-0ubuntu0.12.04.1, generated with red-green.mp4 (https://codereview.chromium.org/8342021):
ffmpeg -i red-green.mp4 -vcodec copy -vbsf h264_mp4toannexb -an red-green.h264
Dumped compressed stream of videos on http://crosvideo.appspot.com manually changing resolutions at random. Those contain 144p, 240p, 360p, 480p, 720p, and 1080p frames. Those frame sizes can be found by
ffprobe -show_frames resolution_change_500frames.vp8
Extract 240 frames using ffmpeg from http://commondatastorage.googleapis.com/chromiumos-test-assets-public/MSE/switch_1080p_720p.mp4.
The frame sizes change between 1080p and 720p every 24 frames.
First 40 raw i420 frames of bear-1280x720.mp4 scaled down to 320x192 for video_encode_accelerator_tests. Encoded with vp9 lossless: ffmpeg -pix_fmt yuv420p -s:v 320x192 -r 30 -i bear_320x192_40frames.yuv -lossless 1 bear_320x192_40frames.yuv.webm
First 40 raw i420 frames of bear-1280x720.mp4 scaled down to 340x384 for video_encode_accelerator_tests. Encoded with vp9 lossless: ffmpeg -pix_fmt yuv420p -s:v 640x384 -r 30 -i bear_640x384_40frames.yuv -lossless 1 bear_640x384_40frames.yuv.webm
First frame of bear_320x192_40frames.yuv for image_processor_test. To get the uncompressed yuv, execute the following command. vpxdec bear_320x192.i420.yuv.webm --rawvideo -o bear_320x192.i420.yuv
Metadata describing bear_320x192.i420.yuv.
First frame of bear_320x192_40frames.yuv for image_processor_test and formatted nv12. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 320x192 -pix_fmt yuv420p -i bear_320x192.i420.yuv -c:v rawvideo -pix_fmt nv12 bear_320x192.nv12.yuv
Metadata describing bear_320x192.nv12.yuv.
First frame of bear_320x192_40frames.yv12.yuv for image_processor_test and formatted yv12. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 320x192 -pix_fmt yuv420p -i bear_320x192.i420.yuv -c:v rawvideo -pix_fmt yuv420p -vf shuffleplanes=0:2:1 bear_320x192.yv12.yuv
RAW RGBA format data. This data is created from bear_320x192.i420.yuv. Alpha channel is always 0xFF. To get the uncompressed yuv, execute the following command. ffmpeg -s 320x192 -pix_fmt yuv420p -i bear_320x192.i420.yuv -vcodec rawvideo -f image2 -pix_fmt rgba bear_320x192.rgba
RAW BGRA format data. This data is created from bear_320x192.i420.yuv. Alpha channel is always 0xFF. To get the uncompressed yuv, execute the following command. ffmpeg -s 320x192 -pix_fmt yuv420p -i bear_320x192.i420.yuv -vcodec rawvideo -f image2 -pix_fmt bgra bear_320x192.bgra
Rotate bear_320x192.nv12.yuv by 90 degrees clockwise. ffmpeg -s:v 320x192 -pix_fmt nv12 -i bear_320x192.nv12.yuv -vf transpose=1 -c:v rawvideo -pix_fmt nv12 bear_192x320_90.nv12.yuv
Metadata describing bear_192x320_90.nv12.yuv
Rotate bear_320x192.nv12.yuv by 180 degrees clockwise. ffmpeg -s:v 320x192 -pix_fmt nv12 -i bear_320x192.nv12.yuv -vf "transpose=2,transpose=2" -c:v rawvideo -pix_fmt nv12 bear_320x192_180.nv12.yuv
Metadata describing bear_320x192_180.nv12.yuv
Rotate bear_320x192.nv12.yuv by 270 degrees clockwise. ffmpeg -s:v 320x192 -pix_fmt nv12 -i bear_320x192.nv12.yuv -vf transpose=2 -c:v rawvideo -pix_fmt nv12 bear_192x320_270.nv12.yuv
Metadata describing bear_192x320_270.nv12.yuv
RAW NV12 format data. The width and height are 1280 and 720, respectively. This data is created from peach_pi-1280x720.jpg by the following command. ffmpeg -i peach_pi-1280x720.jpg -s 1280x720 -pix_fmt nv12 puppets-1280x720.nv12.yuv
To get the uncompressed yuv, execute the following commands. vpxdec puppets-1280x720.i420.yuv.webm --rawvideo -o puppets-1280x720.i420.yuv ffmpeg -s:v 1280x720 -pix_fmt yuv420p -i puppets-1280x720.i420.yuv -c:v rawvideo -pix_fmt nv12 puppets-1280x720.nv12.yuv
RAW NV12 format data. The width and height are 640 and 360, respectively. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 1280x720 -pix_fmt nv12 -i puppets-1280x720.nv12.yuv -vf scale=640x360 -c:v rawvideo -pix_fmt nv12 puppets-640x360.nv12.yuv
RAW NV12 format data. The width and height are 640 and 360, respectively. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 1280x720 -pix_fmt nv12 -i puppets-1280x720.nv12.yuv -vf scale=480x270 -c:v rawvideo -pix_fmt nv12 puppets-480x270.nv12.yuv
RAW NV12 format data. The width and height are 640 and 480, respectively. The meaningful image is at the rectangle (0, 0, 640x360) in the image. The area outside this rectangle is black. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 640x360 -pix_fmt nv12 -i puppets-640x360.nv12.yuv -vf "scale=640:480:force_original_aspect_ratio=decrease,pad=640:480:(ow-iw)/2:(oh-ih)/2" -c:v rawvideo -pix_fmt nv12 puppets-640x360_in_640x480.nv12.yuv
RAW NV12 format data. The width and height are 320 and 180, respectively. To get the uncompressed yuv, execute the following command. ffmpeg -s:v 1280x720 -pix_fmt nv12 -i puppets-1280x720.nv12.yuv -vf scale=320x180 -c:v rawvideo -pix_fmt nv12 puppets-320x180.nv12.yuv
Created using “avconv -i bear-vp9.webm -vcodec copy -an -f ivf bear-vp9.ivf”.
Manually dumped from libvpx with bear-vp9.ivf and test-25fps.vp9. See vp9_parser_unittest.cc for description of their format.
The first frame of blackwhite_yuv444p.mp4 by the following command. ffmpeg -i blackwhite_yuv444p.mp4 -vcodec copy -vframes 1 blackwhite_yuv444p-frame.h264
Used by h265_parser_unittest.cc.
Used by h265_parser_unittest.cc. Copied from bbb_hevc_176x144_176kbps_60fps.hevc in Android repo.
SPS and PPS from bear.hevc for h265_decoder_unittest.cc.
Single IDR, P, B, B, B, P frames respectively from bear.hevc for h265_decoder_unittest.cc.
The first four frames of the HEVC version of bear-av1-320x180-10bit.mp4 created using the following command. ffmpeg -i bear-av1-320x180-10bit.mp4 -vcodec hevc -vframes 4 bear-320x180-10bit-4frames.hevc The file is then split into bitstreams each of which contains a single frame, so that they contain frames as below. bear-320x180-10bit-frame-0.hevc: SPS+PPS+Single IDR bear-320x180-10bit-frame-1.hevc: B bear-320x180-10bit-frame-2.hevc: B bear-320x180-10bit-frame-3.hevc: P
The first frame of blackwhite_yuv444p.mp4 coded in HEVC by the following command. ffmpeg -i blackwhite_yuv444p.mp4 -vcodec hevc -vframes 1 blackwhite_yuv444p-frame.hevc
WebM file containing 12 seconds of solid green video + 300Hz sine wave audio
WebM file containing 10 seconds of solid red video + 500Hz sine wave audio
Created with the following commands:
ffmpeg -f lavfi -i color=c=green:size=160x120 -t 12 -c:v libvpx green.webm ffmpeg -f lavfi -i color=c=red:size=320x240 -t 10 -c:v libvpx red.webm ffmpeg -f lavfi -i "sine=frequency=300:sample_rate=48000" -t 12 -c:v libvpx a300hz.webm ffmpeg -f lavfi -i "sine=frequency=500:sample_rate=48000" -t 10 -c:v libvpx a500hz.webm ffmpeg -i green.webm -i a300hz.webm -map 0 -map 1 green-a300hz.webm ffmpeg -i red.webm -i a500hz.webm -map 0 -map 1 red-a500hz.webm
A lossy WebP encoded image of 2015x2015. Created by gildekel@ using Gimp.
An animated (extended) WebP encoded image of 450x450. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 3000x3000. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 2015x2015 that contains RGB noise. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 115x115 that contains large pixels RGB noise. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 2015x2015 that contains large pixels RGB noise. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 4000x4000 that contains large pixels RGB noise. Created by gildekel@ using Gimp.
A lossy WebP encoded image of 2015x2015 of solid bright green. Created by gildekel@ using Gimp.
A lossless WebP encoded image of 3000x3000. Created by gildekel@ using Gimp.
Single MJPEG encoded frame of 1280x720, captured on Chromebook Pixel. This image does not have Huffman table.
A version of pixel-1280x720.jpg with five trailing zero bytes after the EOI marker. The command used to generated it was:
echo -e "`xxd -g1 -p -c1 pixel-1280x720.jpg`" "\n00\n00\n00\n00\n00" | xxd -r -g1 -p -c1 > pixel-1280x720-trailing-zeros.jpg
A version of pixel-1280x720.jpg converted to grayscale using:
jpegtran -grayscale pixel-1280x720.jpg > pixel-1280x720-grayscale.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
A version of pixel-1280x720.jpg converted to 4:2:0 subsampling using:
convert pixel-1280x720.jpg -sampling-factor 4:2:0 -define jpeg:optimize-coding=false pixel-1280x720-yuv420.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
A version of pixel-1280x720-yuv420.jpg resized to 40x23 (so that the height is odd) using:
convert pixel-1280x720-yuv420.jpg -resize 40x23\! -define jpeg:optimize-coding=false pixel-40x23-yuv420.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
A version of pixel-1280x720-yuv420.jpg resized to 41x22 (so that the width is odd) using:
convert pixel-1280x720-yuv420.jpg -resize 41x22\! -define jpeg:optimize-coding=false pixel-41x22-yuv420.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
A version of pixel-1280x720-yuv420.jpg resized to 41x23 (so that both dimensions are odd) using:
convert pixel-1280x720-yuv420.jpg -resize 41x23\! -define jpeg:optimize-coding=false pixel-41x23-yuv420.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
A version of pixel-1280x720.jpg converted to 4:4:4 subsampling using:
convert pixel-1280x720.jpg -sampling-factor 4:4:4 -define jpeg:optimize-coding=false pixel-1280x720-yuv444.jpg
Then, using a hex editor, the Huffman table sections were removed from the resulting file.
Single MJPEG encoded frame of 1280x720, captured on Samsung Chromebook 2(13"). This image has Huffman table.
1x1 small picture to test special cases.
Size in TKHD is (639.2x360) and size in STSD is (470x360). A PASP box is present with hSpacing=34 and vSpacing=25. Note that 470.0 * 34 / 25 = 639.2.
Size in TKHD is (639.2x360) and size in STSD is (470x360). No PASP box is present.
AC3 audio in framented MP4, generated with
ffmpeg -i bear.ac3 -acodec copy -movflags frag_keyframe bear-ac3-only-frag.mp4
EAC3 audio in framented MP4, generated with
ffmpeg -i bear.eac3 -acodec copy -movflags frag_keyframe bear-eac3-only-frag.mp4
Generated by the following command:
ffmpeg -i bear-1280x720.mp4 -c:v libx264 -c:a libfdk_aac -profile:a aac_he bear-1280x720-aac_he.ts
Generated by the following commands:
ffmpeg -i bear_pcm.wav -c:a mp3 -ar 44100 bear-audio-mp4a.6B.ts
Generated by the following commands:
ffmpeg -i bear_pcm.wav -c:a mp3 -ar 22050 bear-audio-mp4a.69.ts
HEVC video stream in fragmented MP4 container, generated with
ffmpeg -i bear-320x240.webm -c:v libx265 -an -movflags faststart+frag_keyframe bear-320x240-v_frag-hevc.mp4
HEVC video stream in fragmented MP4 container, including the first 2 frames, generated with
ffmpeg -i bear-320x240.webm -c:v libx265 -an -movflags frag_keyframe+empty_moov+default_base_moof \
-vframes 2 bear-320x240-v-2frames_frag-hevc.mp4
This is bear-320x240-v-2frames_frag-hevc.mp4, with manually updated trun.first_sample_flags: s/0x02000000/0x01010000 (first frame is non-sync-sample, depends on another frame, mismatches compressed h265 first frame's keyframe-ness).
This is bear-320x240-v-2frames_frag-hevc.mp4, with manually updated tfhd.default_sample_flags: s/0x01010000/0x02000000 (second frame is sync-sample, doesn‘t depend on other frames, mismatches compressed h265 second frame’s nonkeyframe-ness).
(c) copyright 2008, Blender Foundation / www.bigbuckbunny.org
Generated using following steps:
wget http://distribution.bbb3d.renderfarming.net/video/mp4/bbb_sunflower_1080p_30fps_normal.mp4
ffmpeg -i bbb_sunflower_1080p_30fps_normal.mp4 -c:v libx264 -crf 36 -vf scale=320:240 -c:a libfdk_aac -ac 2 -t 24 bbb1.mp4
ffmpeg -i bbb_sunflower_1080p_30fps_normal.mp4 -f lavfi -i "sine=frequency=500:sample_rate=48000" -map 0:v -map 1:a -c:v libx264 -crf 36 -vf scale=320:240,vflip -c:a libfdk_aac -ac 2 -t 24 bbb2.mp4
ffmpeg -i bbb1.mp4 -i bbb2.mp4 -map 0:0 -map 0:1 -map 1:0 -map 1:1 -c:v copy -c:a copy -movflags frag_keyframe+omit_tfhd_offset+separate_moof bbb-320x240-2video-2audio.mp4
Generated using following commands:
ffmpeg -f lavfi -i color=c=red:size=320x240 -t 5 -c:v libvpx red.webm ffmpeg -f lavfi -i color=c=green:size=320x240 -t 5 -c:v libvpx green.webm ffmpeg -f lavfi -i color=c=blue:size=160x120 -t 10 -c:v libvpx blue.webm ffmpeg -f lavfi -i "sine=frequency=300:sample_rate=48000" -t 10 -c:v libvpx a300hz.webm ffmpeg -f lavfi -i "sine=frequency=500:sample_rate=48000" -t 5 -c:v libvpx a500hz.webm ffmpeg -i red.webm -i green.webm -i blue.webm -i a300hz.webm -i a500hz.webm -map 0 -map 1 -map 2 -map 3 -map 4 multitrack-3video-2audio.webm
bear_silent.webm video injected with “stereo_mode=SIDE_BY_SIDE_LEFT_EYE_FIRST”, “projectionType=EQUIRECTANGULAR”, and projection pose_yaw, pose_pitch, and pose_roll = 10, 20, and 30 respectively.
https://people.xiph.org/~greg/opus_testvectors/