The download preserves the processed channel samples and sample rate in a new PCM16 WAV, but it does not preserve the source codec, bitrate, compression quality, tags, artwork, chapters, cue points, loop markers, or other container metadata.
That makes the download a processed working copy rather than a renamed source or a bit-for-bit copy of the original file. Audio Effects Online keeps the untouched original, performs file reading, decoding, effect processing, WAV encoding, and download creation in the current browser tab, and does not upload the audio or store it in an account. The downloadable result is a new interleaved, little-endian PCM16 RIFF/WAVE file.

What the download actually preserves
The most useful distinction is between audio content and file metadata. Audio Effects Online carries the decoded audio through the selected effect and represents the resulting channel samples and sample rate in the new WAV. It does not carry the original encoded file structure along with them.
| Property | Result in the download | Practical meaning |
|---|---|---|
| Processed channel samples | Preserved in the PCM16 WAV | The effect's blended, reversed, faded, repeated, or normalized sample values are represented after conversion to signed 16-bit PCM. |
| Sample rate | Preserved | The decoded recording keeps its sample rate in the exported WAV. |
| Echo or reverb tail | Included in full when the result fits the output budget | The final delayed samples are not intentionally cut off. |
| Source codec and bitrate | Not preserved | The download is PCM16 WAV rather than a renamed MP3, AAC, Ogg, WebM, FLAC, or other source encoding. |
| Compression quality and artifacts | The encoding is not preserved | Decoded artifacts may remain audible, but the original compressed bits and bitrate are not retained. |
| Tags and container metadata | Not preserved | Artwork, chapters, cue points, loop markers, and other metadata do not carry into the WAV. |
| Untouched source file | Kept separate | The original remains available for archival or a fresh processing run. |
“Preserved” therefore refers to the processed sample content and sample rate, not to every part of the original file. If metadata or the original delivery format matters, keep the source file rather than expecting the effect download to reproduce it.
Why the WAV differs from the source file
The source must first be decoded before an effect can operate on individual samples. The W3C Web Audio specification defines decoded audio in terms of AudioBuffer sample frames, and the browser's Web Audio implementation determines which encoded files can actually be decoded. Once decoded, the chosen effect alters or blends the sample values, finite samples are constrained to the legal −1 through +1 range, and the result is quantized to signed 16-bit PCM.
The downloaded WAV is therefore not the original byte stream. It uses the standard RIFF/WAVE structure and stores the channels as interleaved PCM samples. Microsoft's WAVE format documentation defines block alignment as channels multiplied by bits per sample and divided by eight; for 16-bit PCM, one mono frame occupies two bytes and one stereo frame occupies four.
A source recording may keep the same audible timing and sample rate while gaining an echo, fade, reversal, or a new peak level. Compression artifacts from the original encoding may also remain audible because decoding reconstructs samples rather than restoring the lost compressed data.
Choose one effect and know what changes
Each run applies one selected transformation, and the amount must be a whole number from 0% through 100%. The amount blends or scales the chosen operation rather than invoking a hidden AI transformation.
| Effect | Exact operation | Important limitation |
|---|---|---|
| Echo | At 100%, each input sample is retained and a half-strength copy is added 200 milliseconds later. Lower amounts scale the delayed copy linearly. | Overlapping dry and delayed samples can clip, and the complete tail adds output frames. |
| Lightweight reverb | Copies are added at 29.7, 37.1, and 41.1 milliseconds with disclosed weights of 0.50, 0.35, and 0.25 multiplied by the selected amount. | It is an early-reflection design, not a room, plate, spring, studio, or impulse-response simulation. |
| Reverse | The sample frames are read in the opposite order and blended with the original according to the amount. | The result contains the selected blend, not a permanent reversal of the source file. |
| Fade in | The signal scales from the selected dry proportion at the beginning to full level at the end. | It changes the amplitude envelope rather than the sample rate. |
| Fade out | The fade operates inversely, moving from full level toward the selected dry proportion. | It changes the amplitude envelope rather than trimming the file to a shorter duration. |
| Normalization | The largest absolute decoded sample is used to scale the signal toward a 0.95 sample peak. | It is not LUFS normalization, perceived-loudness matching, dynamics processing, true-peak control, or broadcast compliance. |
The lightweight reverb coefficients are fixed product choices, not measurements of a concert hall or professional processor. Its purpose is a quick sound-design draft with added space or repetition, not a claim of acoustic accuracy.
Apply the effect and download the WAV
- Protect the original and choose the input. Keep a separate untouched copy, then select one browser-decodable encoded audio file no larger than 50 MiB. Accepted filename types include MP3, WAV, M4A, AAC, Ogg, WebM, and FLAC, but a recognized extension does not guarantee that the browser can decode that particular codec and profile.
- Select one effect and amount. Choose echo, lightweight reverb, reverse, fade in, fade out, or normalization, then enter a whole-number amount from 0% through 100%. Only one effect is applied in each run.
- Apply the operation locally. The browser decodes the file, processes its samples, and creates the WAV result in the current tab. If decoding fails, the tool creates no result rather than returning a partial file.
- Inspect the processed data. Review the reported channel count, sample rate, input frame count, output frame count, output peak, and exact WAV byte size. A longer output frame count is expected when echo or reverb adds its calculated tail.
- Download the completed PCM16 WAV. Save the new file separately from the original. If the calculated output would exceed 30 million channel samples, the tool rejects it without truncating the tail and does not create a downloadable file.
- Build a chain only after checking each stage. To use another effect, download the current WAV, inspect it, and select that WAV as the input for a later run. Changing the file, effect, or amount invalidates older work, so reapply and inspect before downloading again.
Limits that can prevent a download
Several limits affect whether processing can finish, and they apply to the encoded input, decoded data, or calculated output rather than to the download format alone.
- Encoded size: The chosen file must be no larger than 50 MiB.
- Actual decodability: MP3, WAV, M4A, AAC, Ogg, WebM, and FLAC names are accepted, but codec and profile support vary by browser and operating system. A filename alone does not guarantee successful decoding.
- Channel and sample-rate range: The decoded input must have 1–8 equal-length channels and a sample rate from 8,000 through 192,000 Hz.
- Decoded input budget: The decoded input may contain no more than 30 million channel samples.
- Complete output budget: The full effect output may contain no more than 30 million channel samples. Echo and reverb can pass the input budget but fail the output budget because their tails add frames.
- No partial fallback: When decoding fails or the complete output exceeds the budget, the tool does not truncate the result or create a file.
For a closer explanation of the distinction between encoded file size and decoded working size, see what the 50 MiB limit means for Audio Effects Online.
Signal level, silence, and chained effects
Strong echo or reverb settings can push overlapping samples beyond the available range. Finite output samples are therefore clipped to −1 through +1 before PCM quantization, which means an output peak at either limit may indicate that clipping occurred. Normalization follows a different rule: it scales toward a 0.95 sample peak based on the largest absolute decoded sample, without measuring LUFS, perceived loudness, equal loudness, or inter-sample true peaks. Silent input remains silent.
A downloaded WAV can serve as the next run's input, but each stage should be inspected. Repeated decoding and PCM processing—especially when the first source was lossy—may make source artifacts more noticeable, while repeated clipping or normalization can change the dynamics. The original file remains the cleanest reference for comparison.
If the goal is to retain an archival master, keep the source encoding and its metadata. If the goal is to use the effect result, the PCM16 WAV preserves the processed channel samples and sample rate in a broadly compatible format while deliberately separating the result from the original container and codec.
If you're weighing options, What Is the Best Way to Use Audio Effects Online? covers this in detail.