WitneyWW's picture
Random sample of 12 from the speech-free + music-free manifest
ca9fb0a verified
Raw
History Blame Contribute Delete
22.7 kB
<!doctype html>
<html lang="en"><head><meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Speech-free &amp; music-free clips &mdash; random sample</title>
<style>
:root{--bg:#0c0d11;--panel:#14161d;--panel2:#1a1d26;--line:#262a36;--fg:#e8eaf0;--dim:#9aa1b2;
--in:#6aa5ff;--ok:#6ee7a8;--warn:#ffd98a;--acc:#8b7bff}
*{box-sizing:border-box} body{margin:0;background:var(--bg);color:var(--fg);
font:15px/1.55 ui-sans-serif,-apple-system,"Segoe UI",Roboto,Inter,sans-serif}
.wrap{max-width:1200px;margin:0 auto;padding:0 20px 80px} a{color:var(--in)}
header.top{padding:44px 0 22px;border-bottom:1px solid var(--line)}
h1{margin:0 0 6px;font-size:29px;letter-spacing:-.02em}
h1 small{font-size:15px;font-weight:400;color:var(--dim)}
.lede{color:var(--dim);max-width:80ch;margin:10px 0 0}
.facts{display:flex;flex-wrap:wrap;gap:10px;margin-top:18px}
.fact{background:var(--panel);border:1px solid var(--line);border-radius:10px;padding:9px 13px;font-size:13px}
.fact b{color:var(--fg)} .fact span{color:var(--dim)}
.note{background:var(--panel);border:1px solid var(--line);border-left:3px solid var(--warn);
border-radius:10px;padding:14px 18px;margin:22px 0;color:var(--dim);font-size:13.5px}
.note b{color:var(--fg)} code{color:var(--acc);font-size:12px}
.card{background:var(--panel);border:1px solid var(--line);border-radius:14px;padding:18px;margin:18px 0}
.chead{display:flex;justify-content:space-between;align-items:baseline;gap:12px;flex-wrap:wrap}
.tags{display:flex;gap:7px;align-items:center;flex-wrap:wrap}
.chip{font-size:12px;padding:3px 10px;border-radius:6px;font-weight:600}
.chip-tag{background:#173a2a;color:#8ff0bd} .chip-warn{background:#453718;color:var(--warn)}
.chip-ds{background:#1b2740;color:#9dc0ff} .chip-sub{background:var(--panel2);color:var(--dim)}
.meta,.pathid{font:11.5px ui-monospace,Menlo,monospace;color:var(--dim)}
.pathid{margin:6px 0 12px}
.body{display:grid;grid-template-columns:minmax(0,1fr) minmax(0,1.15fr);gap:16px}
video{width:100%;border-radius:9px;background:#000;display:block}
figure{margin:0 0 10px} figure img{width:100%;border-radius:7px;display:block}
figcaption{font-size:11px;color:var(--dim);margin-top:4px}
.chips{margin-top:12px;display:flex;flex-wrap:wrap;gap:6px}
.tg{font-size:11.5px;background:var(--panel2);border:1px solid var(--line);border-radius:999px;
padding:3px 10px;color:var(--dim)} .tg b{color:var(--fg);font-variant-numeric:tabular-nums}
.tg.sp{border-color:#5a4a20;color:var(--warn)}
.stats{display:grid;grid-template-columns:repeat(auto-fit,minmax(160px,1fr));gap:10px;margin-top:12px}
.stats div{background:var(--panel2);border:1px solid var(--line);border-radius:9px;padding:9px 12px}
.stats b{display:block;font-size:15px;word-break:break-word}
.stats span{font-size:11px;color:var(--dim)}
footer{color:var(--dim);font-size:13px;border-top:1px solid var(--line);padding-top:20px;margin-top:30px}
@media (max-width:860px){.body{grid-template-columns:1fr}}
</style></head><body><div class="wrap">
<header class="top">
<h1>Speech-free &amp; music-free source clips <small>&mdash; a random sample of 12</small></h1>
<p class="lede">Drawn at random (seed 20260809) from the <b>8,191</b> clips that survived a two-stage
audio filter over 101,503 source videos from InsAVE-80K and JAVEdit-100k: no speech
(FireRedVAD, then FireRedASR2-AED confirmation on every VAD positive) and no music-related
AudioSet tag anywhere in the top five. These are the <i>inputs</i> to the removal pipeline, not its
outputs. Play with sound on.</p>
<div class="facts">
<div class="fact"><b>101,503</b> <span>clips screened</span></div>
<div class="fact"><b>31,845</b> <span>speech-free</span></div>
<div class="fact"><b>8,191</b> <span>also music-free</span></div>
<div class="fact"><b>12</b> <span>sampled at random, not curated</span></div>
</div>
</header>
<div class="note">
<b>Sampled, not cherry-picked</b> &mdash; so the failure mode is visible too. 3 of these 12 carry
<code>Speech</code> as their top AudioSet tag even though VAD and ASR both cleared them. That is the
known model disagreement: FireRedASR2 covers Chinese and English, so speech in another language
transcribes empty and slips through. Those clips are flagged in the manifest via
<code>speech_disagreement</code>, and the stricter variant
(<code>kept_no_speech_strict.csv</code>, 6,983 clips) drops them.
</div>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Sheep</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">general_editing</span></div>
<div class="meta">1280&times;704 &middot; 24 fps &middot;
5.043s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_general_editing_48600</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_general_editing_48600.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_general_editing_48600_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_general_editing_48600_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Sheep <b>0.67</b></span><span class="tg">Bleat <b>0.59</b></span><span class="tg">Animal <b>0.57</b></span><span class="tg">Goat <b>0.55</b></span><span class="tg">Livestock, farm animals, working animals <b>0.20</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0066</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Animal</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">eval</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
3.397s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_eval_00231</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_eval_00231.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_eval_00231_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_eval_00231_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Animal <b>0.12</b></span><span class="tg sp">Speech <b>0.11</b></span><span class="tg">Door <b>0.11</b></span><span class="tg">Bird <b>0.07</b></span><span class="tg">Chicken, rooster <b>0.06</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0161</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Speech</span><span class="chip chip-warn">AudioSet says speech</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">add_and_remove</span></div>
<div class="meta">1280&times;704 &middot; 24 fps &middot;
3.397s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_add_and_remove_04959</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_add_and_remove_04959.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_add_and_remove_04959_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_add_and_remove_04959_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg sp">Speech <b>0.37</b></span><span class="tg">Animal <b>0.07</b></span><span class="tg">Sneeze <b>0.06</b></span><span class="tg">Gasp <b>0.04</b></span><span class="tg">Clip-clop <b>0.03</b></span></div>
<div class="stats">
<div><b>0.96s</b><span>voice activity detected</span></div>
<div><b>yes</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0069</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Speech</span><span class="chip chip-warn">AudioSet says speech</span>
<span class="chip chip-ds">JAVEdit-100k</span>
<span class="chip chip-sub">source</span></div>
<div class="meta">1280&times;720 &middot; 25 fps &middot;
4.864s &middot; aac 16kHz mono</div>
</header>
<div class="pathid">javedit_7162ff62de8f</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/javedit_7162ff62de8f.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/javedit_7162ff62de8f_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/javedit_7162ff62de8f_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg sp">Speech <b>0.81</b></span><span class="tg">Silence <b>0.08</b></span><span class="tg">Television <b>0.06</b></span><span class="tg sp">Male speech, man speaking <b>0.05</b></span><span class="tg">Inside, small room <b>0.03</b></span></div>
<div class="stats">
<div><b>1.23s</b><span>voice activity detected</span></div>
<div><b>yes</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0049</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Speech</span><span class="chip chip-warn">AudioSet says speech</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">general_editing</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
6.715s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_general_editing_34794</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_general_editing_34794.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_general_editing_34794_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_general_editing_34794_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg sp">Speech <b>0.66</b></span><span class="tg">Animal <b>0.28</b></span><span class="tg">Bird <b>0.10</b></span><span class="tg">Duck <b>0.05</b></span><span class="tg">Quack <b>0.05</b></span></div>
<div class="stats">
<div><b>2.18s</b><span>voice activity detected</span></div>
<div><b>yes</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0092</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Door</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">add_and_remove</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
5.069s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_add_and_remove_07472</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_add_and_remove_07472.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_add_and_remove_07472_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_add_and_remove_07472_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Door <b>0.30</b></span><span class="tg">Slam <b>0.17</b></span><span class="tg">Walk, footsteps <b>0.09</b></span><span class="tg">Thunk <b>0.08</b></span><span class="tg">Sliding door <b>0.05</b></span></div>
<div class="stats">
<div><b>0.61s</b><span>voice activity detected</span></div>
<div><b>yes</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0206</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Liquid</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">general_editing</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
3.397s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_general_editing_31165</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_general_editing_31165.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_general_editing_31165_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_general_editing_31165_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Liquid <b>0.15</b></span><span class="tg">Rowboat, canoe, kayak <b>0.11</b></span><span class="tg">Boat, Water vehicle <b>0.09</b></span><span class="tg">Stream <b>0.08</b></span><span class="tg">Animal <b>0.08</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0029</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Frog</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">general_editing</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
3.397s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_general_editing_47918</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_general_editing_47918.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_general_editing_47918_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_general_editing_47918_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Frog <b>0.30</b></span><span class="tg">Croak <b>0.11</b></span><span class="tg">Animal <b>0.05</b></span><span class="tg">Walk, footsteps <b>0.04</b></span><span class="tg">Outside, rural or natural <b>0.04</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0019</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Typewriter</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">add_and_remove</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
6.715s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_add_and_remove_04160</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_add_and_remove_04160.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_add_and_remove_04160_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_add_and_remove_04160_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Typewriter <b>0.28</b></span><span class="tg">Door <b>0.27</b></span><span class="tg">Sliding door <b>0.12</b></span><span class="tg">Printer <b>0.04</b></span><span class="tg">Cash register <b>0.04</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0018</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Clip-clop</span>
<span class="chip chip-ds">JAVEdit-100k</span>
<span class="chip chip-sub">source</span></div>
<div class="meta">1280&times;720 &middot; 25 fps &middot;
4.864s &middot; aac 16kHz mono</div>
</header>
<div class="pathid">javedit_5b6a0a248356</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/javedit_5b6a0a248356.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/javedit_5b6a0a248356_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/javedit_5b6a0a248356_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Clip-clop <b>0.63</b></span><span class="tg">Animal <b>0.57</b></span><span class="tg">Horse <b>0.40</b></span><span class="tg">Walk, footsteps <b>0.10</b></span><span class="tg">Outside, rural or natural <b>0.08</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0024</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Vehicle</span>
<span class="chip chip-ds">JAVEdit-100k</span>
<span class="chip chip-sub">source</span></div>
<div class="meta">1280&times;720 &middot; 25 fps &middot;
4.864s &middot; aac 16kHz mono</div>
</header>
<div class="pathid">javedit_4830e7822b49</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/javedit_4830e7822b49.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/javedit_4830e7822b49_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/javedit_4830e7822b49_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Vehicle <b>0.52</b></span><span class="tg">Car <b>0.13</b></span><span class="tg">Truck <b>0.07</b></span><span class="tg">Medium engine (mid frequency) <b>0.06</b></span><span class="tg">Bus <b>0.05</b></span></div>
<div class="stats">
<div><b>0s</b><span>voice activity detected</span></div>
<div><b>not needed</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0545</b><span>audio RMS</span></div>
</div>
</article>
<article class="card">
<header class="chead">
<div class="tags"><span class="chip chip-tag">Animal</span>
<span class="chip chip-ds">InsAVE-80K</span>
<span class="chip chip-sub">add_and_remove</span></div>
<div class="meta">1280&times;704 &middot; 23.98 fps &middot;
3.397s &middot; mp3 44kHz stereo</div>
</header>
<div class="pathid">insave_add_and_remove_05404</div>
<div class="body">
<div class="vid"><video controls preload="metadata" playsinline src="media/insave_add_and_remove_05404.mp4"></video></div>
<div class="side">
<figure><img loading="lazy" src="media/insave_add_and_remove_05404_strip.jpg" alt="frames">
<figcaption>6 frames, evenly spaced</figcaption></figure>
<figure><img loading="lazy" src="media/insave_add_and_remove_05404_spec.jpg" alt="spectrogram">
<figcaption>audio spectrogram (0&ndash;8 kHz)</figcaption></figure>
</div>
</div>
<div class="chips"><span class="tg">Animal <b>0.39</b></span><span class="tg">Domestic animals, pets <b>0.27</b></span><span class="tg">Bow-wow <b>0.26</b></span><span class="tg">Dog <b>0.21</b></span><span class="tg sp">Speech <b>0.15</b></span></div>
<div class="stats">
<div><b>0.53s</b><span>voice activity detected</span></div>
<div><b>yes</b><span>ASR confirmation</span></div>
<div><b>&mdash;</b><span>transcript (empty = no speech)</span></div>
<div><b>0.0332</b><span>audio RMS</span></div>
</div>
</article>
<footer>
Manifest: <code>/group2/ct/weihanx/JavisDiT/javisdit/script/extra_train.json</code> &mdash; 8,191
entries with dataset attribution, resolvable video paths, AudioSet tags and the per-clip speech
evidence. Clips re-encoded to 360p H.264 / 96&nbsp;kbps AAC for the web; measurements from the
originals.
</footer>
</div>
<script>
document.querySelectorAll('video').forEach(v=>v.addEventListener('play',()=>{
document.querySelectorAll('video').forEach(o=>{if(o!==v)o.pause();});
}));
</script>
</body></html>