Audio Effects Online applies echo, reverb, reverse, fade in, fade out, and normalization to one browser-decodable audio file inside the tab you already have open, then hands you a PCM16 WAV download without sending the audio to any server. The whole pipeline lives in the current browser tab: file reading, Web Audio decoding, sample processing, WAV encoding, and download creation all happen on the device that opened the page. No upload step is required because the audio is never copied off the user's disk. There is also no server-side render and no hidden AI transformation in the background. The trade for that privacy is a narrow, deliberate contract: the tool accepts a single file, applies exactly one disclosed operation to every sample frame, allocates the complete effect tail inside an output limit, and exports a standards-based little-endian PCM16 RIFF/WAVE file. Sticking to those bounds is what lets the effect run in the browser rather than on a remote farm, and it is what makes the result reproducible across runs and machines.
How a Browser Tab Can Process Audio Without an Upload
A web page can do genuine audio work because modern browsers ship with a full audio engine called Web Audio, defined by the W3C Web Audio API specification. That engine exposes an AudioContext, decodes files into in-memory AudioBuffers made of 32-bit float sample frames, and lets JavaScript read or write every individual sample. When you visit a tool that uses Web Audio, the browser is doing the decoding and processing locally; the page does not need to upload a file to a remote service to receive a processed result back.
Audio Effects Online leans entirely on that engine. File reading happens through the standard file input, decoding happens through AudioContext.decodeAudioData, and the only network traffic is the small HTML page itself. Because the same JavaScript runtime that loaded the file is also writing the new sample data and triggering the save dialog, the audio can stay on disk the whole time. After processing the page shows channel count, sample rate, input and output frame counts, output peak, and exact WAV byte size, all derived from the local AudioBuffer. That tight feedback loop is only possible because the audio never leaves the tab.
That local-only contract is the reason the tool is not a generic audio editor with every effect under the sun. It is a fixed, disclosed pipeline: one input, one operation, one output. Picking more than one effect in a single run would require either a heavier in-tab engine or sending samples back and forth, both of which would push the work off the user's device. Keeping the menu small is what lets the same effects run anywhere with nothing leaving the machine.
What the Six Effects Actually Do
The tool offers a short list of effects so each operation is documented, deterministic, and easy to reason about. Picking the right effect starts with understanding what each one does at the sample level, because the same percentage slider behaves very differently across the six options.
Echo is a single deterministic delay. At an amount of one hundred percent, every input sample is retained and a half-strength copy is added exactly two hundred milliseconds later. Lower amounts scale that delayed copy linearly, so at fifty percent only a quarter-strength copy is added, and at zero percent the output equals the input. The full delayed tail is allocated before processing begins, so the last echo is not truncated even on a long input.
Reverb is a disclosed lightweight three-tap early-reflection design rather than a real room or studio emulation. Reflections are added at 29.7, 37.1, and 41.1 milliseconds with weights of 0.50, 0.35, and 0.25 multiplied by the amount slider. These coefficients are stated as product choices, not measurements of any concert hall, plate, spring, studio, impulse response, or branded processor, so the effect does not model room dimensions, frequency-dependent decay, stereo diffusion, damping, pre-delay, convolution, or a professional reverberation tail.
Reverse reads the decoded sample frames in the opposite order and blends them with the original according to the amount, so fifty percent produces a half-volume mirror of the buffer mixed with the original. Fade in scales from the selected dry proportion at the start to full level at the end; fade out performs the inverse. Normalization finds the largest absolute decoded sample and scales toward a 0.95 peak. It is a peak-amplitude target rather than a loudness, LUFS, perceived volume, broadcast compliance, or equal-loudness target, and a silent input stays silent.
| Effect | Operation | Key parameter | Tail or duration |
|---|---|---|---|
| echo | dry + amount × 0.5 × delayed sample | 200 ms fixed delay | full tail allocated |
| reverb | dry + amount × (0.50·t29.7 + 0.35·t37.1 + 0.25·t41.1) | three taps, weights fixed | full tail allocated |
| reverse | (1 − amount) × dry + amount × reversed | 0 to 100 percent blend | same duration as input |
| fade in | linear ramp from amount to 1.0 | full file length | same duration as input |
| fade out | linear ramp from 1.0 to amount | full file length | same duration as input |
| normalization | scale to 0.95 sample peak | peak of input | same duration as input |
Strong echo or reverb settings can clip when dry and delayed samples overlap. Output samples are therefore constrained to the legal −1 through +1 floating range before WAV quantization, so the downloaded file never contains values outside the PCM16 range.
Applying One Effect Without Uploading Your File
- Open Audio Effects Online in a current desktop or mobile browser and choose one audio file no larger than 50 MiB. Recognized filenames include MP3, WAV, M4A, AAC, Ogg, WebM, and FLAC, but actual decoding depends on the browser's Web Audio implementation, so a recognized extension does not guarantee a successful decode.
- Pick one of the six effects from the menu: echo, lightweight reverb, reverse, fade in, fade out, or normalization. The tool applies exactly one effect per run, so choose the operation you need first.
- Set the amount slider to a whole number from 0 to 100. The amount scales or blends each operation as described above; for fade in and fade out it sets the dry proportion at the start or end of the file rather than acting as a simple strength control.
- Apply the effect locally. The page will read the channel count, sample rate, input frame count, output frame count, output peak, and exact WAV byte size after processing. If decoding fails or the output exceeds the channel-sample budget, no file is created.
- Download the PCM16 WAV. The downloaded file is a new interleaved little-endian PCM16 RIFF/WAVE file. Keep the untouched original because the export is a fresh encode, not a wrapped copy of the source.
For a concrete illustration of the budget math, take a 10-second mono recording at 44,100 Hz. The input contains 441,000 sample frames, well under the 30 million channel-sample budget. Picking echo at one hundred percent adds a delayed copy at 200 ms, which is 0.2 × 44,100 = 8,820 sample frames. The output therefore contains 441,000 + 8,820 = 449,820 frames. Encoded as PCM16 mono at two bytes per frame, the output WAV is 449,820 × 2 = 899,640 bytes, while the input frames alone would have produced 882,000 bytes, so the echo adds exactly 17,640 bytes. That arithmetic runs entirely on the local AudioBuffer, which is why the page can show exact byte counts the moment processing finishes: the numbers come straight from the same budget the processor uses.
Limits That Decide Whether the Job Finishes
The local pipeline has explicit boundaries that exist for a reason: they keep the work inside the memory of one browser tab and prevent partial results from being silently written out. Every limit is enforced before the effect runs, and a failure is communicated by simply not producing a download.
| Limit | Value | What happens at the boundary |
|---|---|---|
| File size | up to 50 MiB | Larger files are rejected before decoding |
| Filename recognition | mp3, wav, m4a, aac, ogg, webm, flac | Browser codec support still varies |
| Channels | 1 to 8 equal-length | Outside range, no result is produced |
| Sample rate | 8,000 to 192,000 Hz | Outside range, no result is produced |
| Effect amount | integer 0 to 100 | Other values are rejected |
| Input channel samples | up to 30,000,000 | Over budget, decoding is rejected |
| Output channel samples | up to 30,000,000 | Over budget, nothing is truncated, no file is created |
| Echo and reverb tails | extra frames added | May pass input budget yet fail output budget |
The input and output budgets are checked separately because echo and reverb add frames. A 30-million-sample input can pass the input gate, yet the same file with echo or reverb at full amount can fail the output gate because the tail pushes the total past 30 million channel samples. In that case nothing is truncated: no half-finished WAV is written, no warning dialog is shown, and the only indication is the absence of a download. Stale results cannot replace a newer state either, because each job gets its own identity and the previous object URLs are revoked.
What the Output Keeps and What It Drops
The downloaded PCM16 WAV keeps the processed sample data and the sample rate, but it is a fresh encode rather than a container copy. That distinction matters if your source file carries information beyond the audio waveform. Tags, artwork, chapters, cue points, loop markers, and other container metadata are not preserved, and the source codec, bitrate, and compression quality are not carried over because the audio has been fully decoded to 32-bit float samples and then re-quantized to signed 16-bit PCM.
That fresh-encode behavior is also why the tool exposes the exact output byte size after processing. The WAVE block alignment that Microsoft documents for PCM data is channels multiplied by bits per sample divided by eight, so sixteen-bit mono produces two bytes per frame and sixteen-bit stereo produces four bytes per frame. That relationship, combined with the frame count reported on the page, is what determines the final file size. If you re-encode the same WAV through a lossy codec and then load that codec back in, repeated decode and PCM processing can reveal source artifacts, so it is good practice to keep an untouched original and to use the downloaded WAV as the next input only when you intentionally want a clean chain.
Build a Multi-Effect Chain by Re-Using the WAV
The tool applies only one effect per run by design. To stack effects, download the result, inspect the page's reported frame count and peak, and use that WAV as the next input. Each run is independent: changing the file, the effect, or the amount invalidates older work, so two open runs of the same source with different settings do not interfere with each other.
A reasonable chain is fade in, then normalization, then a light reverb. Apply fade in to remove a hard start, download the WAV, reopen that WAV, and apply normalization to bring the faded-in signal to a 0.95 peak, then download, re-open, and add a small amount of reverb to put the recording in a virtual room. None of those intermediate files leave the device, so the chain stays as private as a single run. If you want to A/B the chain against the untouched original, just keep the original file open in another tab and switch back to compare, since the tool never edits the source.
For readers who want a deeper walkthrough of the local pipeline and the WAV export, the guide on applying audio effects online locally and downloading a WAV covers the same procedure with extra detail. For the underlying decoder contract, the W3C Web Audio API 1.1 specification defines how AudioBuffer length is measured in sample frames and how decodeAudioData behaves.
For a deeper look, see Audio Equalizer Online: Local Effects to Adjust Your Audio.