Original files for a repeatable comparison
Download 21 original synthesized examples for format support, channel checks, loudness, fades, and gating. Choose 16-bit or 24-bit integer PCM and 32-bit float WAVs; the actual rate, encoding, channels, duration, and bytes appear on each file. No account, payment, or audio upload is needed.
Start at a comfortable playback volume. Press play for a manual preview or download a WAV for your editor. Check the described signal rather than treating every silent output as a broken file. These are controlled learning examples, not recordings of speech or proof of AI separation quality.
Compare stored formats with the 16-bit examples in the groups below. These one-second 440 Hz tones have 20 ms edge fades: one is 24-bit integer PCM at 44.1 kHz and the other is 32-bit IEEE float at 48 kHz. They test file handling, not whether a recording sounds better. Browser playback support can differ; inspect the actual header or use a compatible editor. Inspect real WAV formats and compare the header values.
Integer PCM tone for checking format support and stored sample properties. 44 bytes are outside the audio payload. A decoded player may convert the signal internally; playback alone does not verify the stored format.
1 second · Mono · 44.1 kHz · 24-bit PCM
132,344 bytes
IEEE floating point tone for checking format support and stored sample properties. 58 bytes are outside the audio payload. A decoded player may convert the signal internally; playback alone does not verify the stored format.
1 second · Mono · 48 kHz · 32-bit IEEE float
192,058 bytes
Compare matching channels, a left-only source, and equal opposite-polarity channels. Each case includes the original stereo WAV and separate left, right, and averaged-mono exports. A silent right channel or mix is intentional where described; check your playback routing before interpreting it. Read the channel arithmetic and listening method.
The same tone is stored on both sides.
3 seconds · Stereo · 48 kHz · 16-bit PCM
576,044 bytes
The original left tone, played as a one-channel file.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The same tone from the original right channel.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The tone retains its original sample amplitude.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The tone is on the left; the right side is silent.
3 seconds · Stereo · 48 kHz · 16-bit PCM
576,044 bytes
The wanted tone keeps its original sample amplitude.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
Silence is expected: this source channel is empty.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The tone has half the left export’s sample amplitude.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The two channels hold equal tones with opposite polarity.
3 seconds · Stereo · 48 kHz · 16-bit PCM
576,044 bytes
The tone remains audible in the separate left channel.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The opposite-polarity tone remains audible on its own.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
Silence is expected: every averaged sample is zero.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
These four-second 440 Hz recordings both peak at −12.04 dBFS. The steady tone and quieter pulsed pattern have different integrated LUFS. Use complete files to compare peak normalization with a requested loudness target; these synthesized checks do not establish music or speech quality. Read the loudness measurement and normalization comparison.
Generation, verification, and CC0 notes for this group.
The tone keeps its level through the middle, with short fades at the outer edges. Measured integrated loudness: -15.74 LUFS.
4 seconds · Mono · 48 kHz · 16-bit PCM
384,044 bytes
Short louder pulses reach the same peak. Most of each half-second cycle stays at one quarter of that amplitude. Measured integrated loudness: -22.65 LUFS.
4 seconds · Mono · 48 kHz · 16-bit PCM
384,044 bytes
The same three-second 330 Hz source has abrupt edges, 5 ms fades, or a longer 0.5-second entrance and one-second ending. Compare boundaries and preserved duration. These outer-edge examples do not represent repair of a click inside a recording. Read how to locate and treat an edit click.
Generation, verification, and CC0 notes for this group.
The tone starts and stops at a nonzero sample level. Listen to its entrance and ending; no fade is applied.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
A short fade changes only 240 frames at each edge. The middle keeps the original decoded level.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
The tone has a slower entrance and ending. Much more of its duration receives a reduced gain, even though its highest sample peak is unchanged.
3 seconds · Mono · 48 kHz · 16-bit PCM
288,044 bytes
A quiet high tone accompanies two louder low-tone bursts. Compare the original with an actual gated export: the quieter gap is reduced while the competing tone remains during the steady overlap. This is an illustration of a gate's limit, not an AI isolation or speech-cleanup result. Read the gate settings and measured comparison.
Generation, verification, and CC0 notes for this group.
A quiet high tone continues throughout. Two louder low-tone bursts begin at 0.5 and 2 seconds. Listen to the high tone both between and during those bursts.
4 seconds · Mono · 48 kHz · 16-bit PCM
384,044 bytes
The quiet gaps are reduced, but the high tone still passes while a louder burst opens the gate. Quieter pauses do not prove that overlapping background sound was separated.
4 seconds · Mono · 48 kHz · 16-bit PCM
384,078 bytes
Use the files in a real workflow
Open an original stereo file in the channel splitter to compare left, right, and averaged mono. Inspect a downloaded file's channel count and stored format with the WAV inspector. An opposite-polarity source can cancel when averaged even though each channel contains a tone.
Use the complete equal-peak files in the LUFS meter, peak normalizer, or LUFS normalizer. Compare the actual exported measurements and preserve the originals. A requested LUFS target can be constrained by sample-peak headroom.
For an outer-edge experiment, download the abrupt fade source and choose it in the fade tool. The file's middle and timing need a different check from its entrance and ending. Files are not transferred between pages automatically; download and choose the working copy yourself.
Read the scope before comparing
They provide known samples for the specific digital operations described. Your player or device can downmix, apply balance settings, normalize playback, or process the signal further. A listening impression alone does not identify which part of that playback path changed the audio.
The files do not demonstrate a clean separation of real voices or instruments, identify a recording, or certify an editor's complete output behavior. Use representative recordings and suitable final-delivery measurements for those tasks. Some silent files here are deliberate mathematical outcomes, including an empty right channel and an opposite-polarity averaged mix.
Every download is a complete short WAV. Original generation methods and existing file notices are linked by group. The fade, loudness, and gate examples include published CC0 notices; no new blanket reuse license is assigned to the channel or format examples. Follow those group links for generation and reuse details.