Audio Fade In and Fade Out

The Audio Trimmer and Fade Editor prepares a local PCM WAV from one audio file. It offers manual trimming, linear fades, and gain, not loudness normalization, AI cleanup, or automatic music selection.

Decide what the musical entrance needs to preserve

Imagine a local instrumental workshop with a twenty six second recording of a short plucked motif and a ringing final chord. The organizer wants a twelve second sting between two sections of a workshop video. The first usable attack occurs at source time 2.650; the final chord is struck at 11.500 and then decays. Choose a candidate from 2.000 to 14.000 seconds. That gives the opening a little space and lets the final chord ring. These are invented musical timing choices for one supplied recording, not a formula for how every sting should begin or end.

The entrance has a particular job: let the texture arrive without blunting the first pluck. An initial Fade-in of 0.500 seconds finishes before the attack at 2.650, so it mainly shapes the preceding room sound. The proposed Fade-out of 2.000 seconds begins at source time 12.000, after the final chord has been struck. It therefore reduces part of the decay rather than the initial chord attack. Whether that sounds appropriate still requires listening. A different performance with an earlier attack or a longer musical resolution would need a different interval, even if the same twelve second slot were available.

Keep the musical purpose separate from a decorative preference for smooth curves. An abrupt intentional pluck may belong in the piece, while a cut through its resonance may sound accidental. A fade is not automatically an improvement just because the waveform becomes narrower near the edge. Compare what the workshop sequence needs with what this recording can provide. The editor does not find a beat, infer a phrase ending, select an instrumental track, or compose an extra resolving chord. If the chosen material never reaches a suitable ending, choose another performance or make the musical edit in a more appropriate arrangement workflow.

Prepare the source without inventing extra detail

Load one local audio-only file within 32 MiB and 180 seconds, with mono or stereo channels. Supported candidates include WAV, MP3, M4A, OGG, FLAC, and audio-only WebM when the native browser decoder supports their actual codec. A video recording of the workshop is not accepted as an audio extraction source. Nor can you paste a media URL or record an instrument through this page. Bring an existing audio export from the recording workflow. Keep the original performance separately because a short faded WAV is a new asset, not a replacement archive for the full workshop take.

The browser prepares accepted audio at 48000 Hz. A 44100 Hz source may therefore be resampled rather than retaining its original sample grid. That does not create new musical information, repair lossy compression, or make a poor microphone placement more detailed. One source channel stays one and two stay two. The fade and gain operate on the supplied signal, not on separately identified instruments. If a chair scrape overlaps the ringing chord, the envelope reduces both together. There is no denoising, stem separation, pitch change, or speed control that could isolate the instrument and leave the scrape behind.

Listen to the source before entering a fade duration. Check the attack, the final chord, and the background that becomes exposed as the note decays. After a new valid file is accepted, the default selection is its first thirty seconds or its complete duration if shorter. Our twenty six second source therefore begins as a whole-source candidate, not the proposed twelve second sting. Change the range deliberately. The original player and prepared player do not autoplay. This is useful for controlled comparison, but it means that loading a file alone has not auditioned it or confirmed that the intended performance was selected.

Map the selected time range onto sample frames

Set Start time to 2.000 and End time to 14.000. At 48000 frames per second, the first selected frame index is 96000 and the end index is 672000. The end is exclusive, so the selected frames stop just before that end index. The result contains 576000 frames, exactly twelve seconds. Stereo has two channel samples per frame, not twice as many seconds. This distinction matters when describing duration and estimating the WAV size. The fade will be applied within these selected output frames, not to the earlier room sound that you removed from the source.

For general input, the start frame is ceil(start seconds * 48000). The requested end is converted with the same ceiling operation and bounded by the actual prepared source length. A blank End time means the prepared source end, not the end of the last musical phrase. The fields accept milliseconds; the sample grid resolves how those requested values become an actual interval. Read the prepared result for the actual duration rather than assuming that typed decimals describe an unlimited continuous timeline. This is explicit sample addressing, not beat snapping, transient detection, or a guarantee of a musically meaningful boundary.

The selected output must be 0.25 to 120 seconds and remain within the source. An invalid range is not silently corrected into a different musical choice. If you reverse the boundaries or import settings for a longer source, fix the selection before preparing. Once the arithmetic is valid, audition both ends. The waveform represents real decoded PCM but does not tell you whether a chord has resolved or whether the first quiet shape is a useful lead. Mathematical precision makes the chosen edit reproducible; it does not make the choice correct for the workshop sequence without an audible and musical review.

Understand the linear envelope sample by sample

Set Fade-in to 0.500 and Fade-out to 2.000 seconds. Fade frame counts use round(seconds * 48000), giving 24000 opening frames and 96000 closing frames. Both fades must fit without overlapping, and each duration can be zero to ten seconds. Their combined 2.500 seconds fits inside this twelve second selection. A zero fade means that boundary is not faded; a nonzero fade must contain enough frames to have distinct endpoints. The tool rejects a one-frame nonzero fade. These checks define a valid envelope, not an instruction to use the longest fade that the selection happens to permit.

Let G be the constant amplitude gain chosen by Level. For an opening region containing F frames, number those frames n from zero through F minus one. Its ideal linear multiplier is G * n / (F - 1). For the closing region, using a fresh local n across its F frames, it is G * (1 - n / (F - 1)). The first affected opening sample is zero and the last reaches G; the closing region goes from G to zero. Between these regions, the multiplier stays at G. This describes amplitude shaping, not a linear change in perceived loudness.

With Level at -3 dB, G is approximately 0.707946 because the gain factor is 10^(dB / 20). Near the halfway point of either linear ramp, the multiplier is about 0.353973. The envelope is near half its plateau amplitude there; that does not mean listeners hear half the loudness. For this example the opening region ends just before output time 0.500, while the closing region occupies output time 10.000 through the final sample. Those sample endpoint details explain how the render is shaped. They do not remove the need to check whether the attack and decay still feel appropriate in the actual music.

Use gain for headroom, not for repair

The Level control ranges from -24 to +12 dB in 0.1 dB steps. Outside the fades, it applies a constant multiplier to the selected signal. This is not a compressor, limiter, or loudness normalization process. A quiet background and a loud pluck keep their relative relationship under that gain. For the workshop sting, -3 dB is a candidate that reduces the whole signal before it meets the neighboring video sections. It is not a published standard or a universally safe delivery target. Listen at a fixed playback setting when comparing candidates so a change in speaker volume does not confuse the judgement.

Consider a hypothetical source sample with magnitude 0.800 in the unaffected middle of the clip. A +6 dB setting multiplies it by about 1.995262, producing about 1.596210, beyond full scale. The tool refuses a render whose processed floating point peak exceeds 1 instead of silently clipping the WAV. At -3 dB, that same illustrative sample becomes about 0.566357. These are worked arithmetic examples, not measurements of a file that has not been supplied. The actual result must be rendered and checked. A peak elsewhere in the selected material can determine whether a seemingly modest positive setting is accepted.

Lower gain cannot reconstruct a recording that was already clipped. If the original instrument attack has a flattened, distorted shape, reducing it produces a quieter version of that damage. The fade also cannot repair overload in the middle of the sting or remove a scrape beneath the chord. Keep any refusal from adding overload separate from the quality of the existing recording. If distortion is audible before processing, return to another take or the original recording setup when possible. Do not label a lower sample peak as restored dynamics, repaired audio, denoising, or evidence that the instrument has become cleaner.

Audition the join in the full workshop sequence

Press Prepare audio and play the result from its beginning, not only from the loud central phrase. Listen for whether the room sound enters naturally and whether the first pluck still has its intended attack. Then listen through the final chord and the complete two second release. A long fade can remove a musically useful resonance, while a short one can sound like a door closing on it. The numbers give you a controlled experiment. Keep one approved candidate on disk before trying another, and compare equivalent listening conditions rather than relying on a memory of how smooth the previous version seemed.

Import the approved WAV into the video editor and audition its join with the preceding and following sections. A sting that sounds complete on its own can still interrupt the pace of spoken workshop material. Check whether the preceding speaker has finished, whether the instrumental attack surprises the listener, and whether the following voice begins against too much remaining chord. This page processes one source only. It does not crossfade two clips, mix narration, or duck music beneath a voice. If you need an overlapping transition, design and review that mix in the separate timeline rather than assuming two boundary fades create it automatically.

Inspect the finished video after assembly because other processing can change the result. A duplicated audio track, an editor gain setting, or later encoding can undermine an otherwise deliberate sting. If the completed video needs separate file preparation, the Video Compressor handles a different video workflow; it is not the musical fade operation described here. Recheck sound after that step instead of assuming a valid standalone WAV proves the final sequence. The practical approval question is whether the transition supports the workshop pace without masking either neighboring voice, not whether every boundary has the same fade duration.

Interpret the WAV and its actual sample measurements

The production output is 48000 Hz, 16-bit PCM WAV with the source mono or stereo layout. The tool does not export MP3 or AAC. A twelve second stereo version has 576000 frames and 2304000 bytes of sample data, before the small WAV container overhead. The output ceiling is 24 MiB. A trimmed PCM WAV can be larger than a compressed source because its storage model is different, not because a fade adds musical detail. Use this file as the audio asset for the workshop project, and keep the original recording separately if you may need another phrase or a less shortened decay.

Measurements are taken after the actual WAV is re-decoded. Sample peak identifies the largest sample magnitude, while RMS describes the root mean square of the measured sample values. In simplified terms, RMS squares the values, averages them, and takes the square root; reporting that level in dBFS uses a logarithmic scale. A silent result has no finite level to report, so silence may appear as an unknown or null measurement rather than a useful numeric target. The report also describes actual frames, duration, channels, and bytes. These are properties of the prepared file, not predictions based only on your requested settings.

Neither sample peak nor RMS is LUFS or a true peak measurement. They do not certify a streaming destination, a workshop sound system, or a final video mix. Longer fades can reduce RMS by attenuating more samples even when the central musical phrase remains unchanged. Compare candidates with the same range when investigating that effect, and listen rather than choosing the smallest or largest value automatically. If you require a particular measured loudness delivery workflow, use suitable separate measurement and processing. Do not rename this manual gain control normalization or treat a successful export as proof of compliance with an unrelated platform standard.

Preserve a recipe without confusing it with the recording

Download the approved WAV after the final preparation. Any input change makes the existing prepared result stale, so changing a fade from 2.000 to 1.500 seconds requires a new valid render before the current exports are available again. A previously downloaded file on disk does not update itself. Keep your own revision names for the sting and its place in the video project. Save settings JSON for the range, fades, gain, context, authored title, and note. It contains no audio, source filename, or metrics, and therefore cannot reconstruct the instrument performance without the separate matching source.

Opening settings retains the audio currently loaded; it does not prove that the recording matches the recipe. Resolve any invalid range and audition again after importing. For review, use copied text, TXT, CSV, or a styled PDF. The PDF attaches settings and review JSON but no audio or source filename, and its pages are rasterized. Titles and notes remain visible report content, so review them before sharing workshop details. A collaborator assembling the video needs the approved WAV separately. Someone reviewing the transition needs the assembled sequence, because a report about a twelve second asset cannot demonstrate how it joins the neighboring material.

Processing the input is local and there is no autosave, though normal website assets and a public example can load. Leaving the page releases decoded buffers and the prepared WAV. Browser history may retain a File for explicit Restore audio, while a reload requires selecting the source again. Preserve the files you need before leaving. Reopen the Audio Trimmer and Fade Editor for another phrase with a fresh listening decision. If supported account routines later become part of distributing the finished workshop video, SMTasker is separate: this editor does not import settings there or publish the sting as an audio post.

Questions about Audio Fades

Continue with another audio-editing guide

Apply the idea in the Audio Trimmer & Fade Editor, or continue with another focused audio guide. Browse more utilities in all free tools.