Open-source AI browser agent for Chrome and Firefox (monorepo) ðŸ§
JavaScript ai-agentai-agentsai-toolsbrowser-aibrowser-automationbrowser-extensionchrome-aichrome-devtoolschrome-extensionchrome-extensionsfirefox-addonfirefox-addonsfirefox-extensionfirefox-extensionsgemmagemma4minimaxoffline-aioffline-llmqwen
Video fixtures
One-second 64×64 test clips for test/generative-media-video.mjs, made with
FFmpeg 6.1 from its built-in test sources:
| Files | Video | Audio | Container |
|---|---|---|---|
h264-aac-1.mp4, h264-aac-2.mp4 |
H.264, 12 fps, 2 B-frames | AAC, 16 kHz mono | MP4, faststart, edit lists |
h264-aac-fragmented-1.mp4, h264-aac-fragmented-2.mp4 |
H.264, 12 fps, 2 B-frames | AAC, 16 kHz mono | Fragmented MP4, no edit lists |
h264-aac-48px.mp4 |
H.264 at 48×48 | AAC | MP4 (codec setup differs from the others) |
vp9-opus-1.mp4, vp9-opus-2.mp4 |
VP9, 12 fps | Opus, 48 kHz mono | MP4 (royalty-free codecs, playable in Playwright's Chromium) |
h264-aac-short-audio.mp4 |
H.264, 12 fps, 2 B-frames | AAC, 16 kHz mono, ends 0.25 s before the picture | MP4, faststart, edit lists |
h264-aac-late-audio.mp4 |
H.264, 12 fps, 2 B-frames | AAC, 16 kHz mono, starts 0.25 s in (an empty edit) | MP4, faststart, edit lists |
vp9-opus-late-audio.mp4 |
VP9, 12 fps | Opus, 48 kHz mono, starts 0.25 s in (an empty edit) | MP4 |
E="-c:v libx264 -preset veryfast -pix_fmt yuv420p -g 12 -bf 2 -c:a aac -b:a 24k -ac 1"
H="$E -shortest"
ffmpeg -f lavfi -i "testsrc=size=64x64:rate=12:duration=1" -f lavfi -i "sine=frequency=300:sample_rate=16000:duration=1" \
$H -movflags +faststart -fflags +bitexact -map_metadata -1 h264-aac-1.mp4
ffmpeg -f lavfi -i "testsrc=size=64x64:rate=12:duration=1" -f lavfi -i "sine=frequency=500:sample_rate=16000:duration=1" \
$H -movflags frag_keyframe+empty_moov+default_base_moof -fflags +bitexact -map_metadata -1 h264-aac-fragmented-1.mp4
ffmpeg -f lavfi -i "testsrc2=size=64x64:rate=12:duration=1" -f lavfi -i "sine=frequency=400:sample_rate=48000:duration=1" \
-c:v libvpx-vp9 -b:v 60k -g 12 -deadline realtime -cpu-used 8 -c:a libopus -b:a 16k -ac 1 -shortest \
-fflags +bitexact -map_metadata -1 vp9-opus-1.mp4
ffmpeg -f lavfi -i "testsrc=size=48x48:rate=12:duration=1" -f lavfi -i "sine=frequency=300:sample_rate=16000:duration=1" \
$H -movflags +faststart -fflags +bitexact -map_metadata -1 h264-aac-48px.mp4
# Sound shorter than the picture, or starting after it: the picture keeps its full second.
ffmpeg -f lavfi -i "testsrc=size=64x64:rate=12:duration=1" -f lavfi -i "sine=frequency=900:sample_rate=16000:duration=0.75" \
$E -movflags +faststart -fflags +bitexact -map_metadata -1 h264-aac-short-audio.mp4
ffmpeg -f lavfi -i "testsrc=size=64x64:rate=12:duration=1" -itsoffset 0.25 -f lavfi -i "sine=frequency=1100:sample_rate=16000:duration=0.75" \
$E -movflags +faststart -fflags +bitexact -map_metadata -1 h264-aac-late-audio.mp4
ffmpeg -f lavfi -i "testsrc2=size=64x64:rate=12:duration=1" -itsoffset 0.25 -f lavfi -i "sine=frequency=1300:sample_rate=48000:duration=0.75" \
-c:v libvpx-vp9 -b:v 60k -g 12 -deadline realtime -cpu-used 8 -c:a libopus -b:a 16k -ac 1 \
-fflags +bitexact -map_metadata -1 vp9-opus-late-audio.mp4
The second clip of each pair uses a different tone (600, 1000 and 800 Hz), so its bytes and bitrate fields differ while the codec setup matches.