An audio waveform generator is not a spectrogram and it is not a frequency analyzer. It is a time-domain visualization tool: it draws the minimum and maximum sample amplitude for each small slice of audio along a horizontal time axis, producing a static peak envelope image rather than a frequency-versus-time map. A spectrogram requires a Fourier transform to break audio into frequency bins and shows how spectral content changes over time using color or intensity, while a frequency analyzer reports the dominant frequencies, spectrum bands, or pitch information present in a signal. The Audio Waveform Generator from Lizely does neither of those things. It decodes the audio in your current browser tab, divides the decoded samples into evenly spaced buckets, finds the highest and lowest finite sample value across all channels in each bucket, and renders one vertical line per bucket as a scalable SVG file. Because the vertical axis encodes sample amplitude rather than frequency, the resulting image cannot tell you which notes are playing, where formants sit, or how the spectrum evolves over time.
What an audio waveform generator actually shows
The horizontal axis is time, with the first decoded frame on the left and the last decoded frame on the right. The vertical axis is sample amplitude scaled so that full-scale positive and negative values reach forty-five percent of the requested image height above and below the center line, leaving a five-percent margin at the top and bottom of the SVG. Each vertical line represents one time bucket, and the line stretches from the lowest sample in that bucket to the highest. This method preserves short positive and negative peaks more honestly than picking one arbitrary sample per bucket, because transient spikes and brief excursions survive the peak-envelope step instead of disappearing between samples.
The columns in the image are evenly spaced across the requested width. The number of columns is the smaller of three values: the requested image width in pixels, 1,000, and the total decoded frame count. A short clip therefore never invents more independent measurements than it contains, while a long clip stays within a bounded SVG file size and processing budget. Bucket boundaries are computed with integer frame indices, the first bucket begins at frame zero, the last bucket reaches the final decoded frame, and no accepted frames are omitted between adjacent ranges. This makes the result deterministic: the same input file with the same width and height settings always produces the same SVG, down to the exact line coordinates.
Why a waveform generator is not a spectrogram
A spectrogram answers the question "what frequencies are present at this moment?" by applying a short-time Fourier transform to small overlapping windows of audio and plotting the resulting magnitude against time. The horizontal axis is still time, but the vertical axis is frequency, and color or grayscale intensity encodes the magnitude at each time-frequency bin. That requires a Fourier transform, windowing functions, overlap, and careful scaling, none of which the peak envelope method performs.
The Lizely waveform tool works entirely in the time domain. It reads each decoded sample value, sorts those values into time buckets, and reports the minimum and maximum range within each bucket. Nothing about the picture changes if the underlying audio contains strong 60 Hz hum, sharp 10 kHz harmonics, or a slowly rising 200 Hz tone, because those phenomena live in the frequency domain and the tool never converts samples into frequency bins. If you need to see harmonics, formants, or noise profiles, you need a tool that explicitly computes a spectrogram rather than a peak envelope.
How a waveform generator differs from a frequency analyzer
A frequency analyzer takes an audio signal and reports its frequency content, usually as a list of band magnitudes, peak hold values, or real-time bars on a display. Spectrum analyzer hardware and software both fall into this category. They typically use a fast Fourier transform, a bank of bandpass filters, or both, and they answer questions like "what is the strongest frequency right now?" and "how much energy sits between 1 kHz and 2 kHz?" Those readings are numbers on a scale, not a single shape made of vertical lines.
A peak waveform answers a different question: "what is the loudest and quietest sample in each short slice of time?" That information is useful for editing, layout, and quick visual reference, but it carries no frequency data at all. Two recordings can produce nearly identical peak envelopes while sounding completely different, because perceived loudness, pitch, and timbre depend on duration, frequency balance, dynamics, and playback conditions. The peak picture is not a calibrated loudness meter, and treating it that way will produce misleading conclusions about mastering compliance or perceived volume.
When a peak waveform is the right choice
A peak envelope image is well suited for several everyday tasks: checking whether a recording contains actual audio or silence, finding the rough location of loud moments in a long interview, deciding where to trim a track in a video editor, illustrating a podcast episode page, or producing a quick thumbnail that represents a sound. Because the result is a vector SVG, it stays sharp when scaled and can be opened in any SVG-aware editor such as Illustrator, Inkscape, or Affinity Designer without losing detail.
A peak waveform is the wrong choice when you need to identify a musical key, judge room acoustics, diagnose a hearing concern, or check whether a master meets broadcasting loudness targets. None of those tasks can be answered from amplitude peaks alone. They require frequency analysis, calibrated metering, or perceptual models that this tool does not implement.
| Aspect | Peak waveform generator | Spectrogram | Frequency analyzer |
|---|---|---|---|
| Horizontal axis | Time | Time | Frequency or time depending on mode |
| Vertical axis | Sample amplitude from -1 to 1 | Frequency in Hz | Frequency in Hz or magnitude |
| Math applied | Min and max per time bucket | Short-time Fourier transform | FFT or filter-bank analysis |
| Typical output | Static SVG image | Color-coded time-frequency map | Numeric spectrum or bar display |
| Reveals pitch or notes | No | Indirectly via harmonic patterns | Yes with calibration |
| Calibrated loudness meter | No | No | No unless LUFS integrated |
| Works fully offline in browser | Yes with the Lizely tool | Rare for high-resolution versions | Some browser implementations |
How to generate a local peak waveform with the Audio Waveform Generator
- Open the Audio Waveform Generator in your current browser tab and choose one file that the browser can actually decode. Supported formats include MP3, WAV, M4A, AAC, Ogg, WebM, and FLAC, but only when the current browser and operating system can supply the matching codec. The compressed file must be at most 50 MiB, the decoded audio must last no more than five minutes, contain one through eight channels, use a sample rate from 8,000 through 192,000 Hz, and stay within 30 million channel samples overall.
- Set the SVG width between 320 and 1,600 pixels and the height between 120 and 600 pixels, then choose a six-digit hexadecimal background color and a six-digit hexadecimal waveform color from the browser color controls.
- Generate the local peak envelope. The tool divides all decoded frames into contiguous integer-index buckets whose count is the smaller of the requested width, 1,000, and the decoded frame count. For each bucket, it finds the minimum and maximum finite sample across every channel, clamps any defensive out-of-range values to between -1 and 1, and draws one vertical SVG line from the bucket maximum to its minimum.
- Inspect the reported frame count and peak column count, look at the preview, and open the expandable SVG text inspector to confirm dimensions, line count, and coordinates. The download uses that identical string, so what you see in the inspector is exactly what you save.
- Download the SVG. The file is plain text containing one background rectangle and a group of peak lines with a matching viewBox. Open it in an SVG-aware editor, drop it into a layout, or scale it on the web without quality loss.
A concrete example makes the bucket math concrete. Suppose your file decodes to 882,000 frames of stereo audio at 44,100 Hz and you request a 1,200-pixel-wide SVG. The column count is min(1,200, 1,000, 882,000), which equals 1,000. The first bucket covers frames 0 through 881 inclusive, the second bucket covers frames 882 through 1,763, and so on through the 1,000th bucket, which reaches the final decoded frame. Within bucket 1, the tool scans all samples from frame 0 to frame 881 across both channels, picks the lowest finite value, picks the highest finite value, and draws one vertical line spanning the difference. That procedure is repeated for every other bucket.
Limits and what the picture cannot tell you
The peak envelope view deliberately avoids claims it cannot back up. It does not calculate frequency content, beats, notes, loudness, speech, transients, or musical structure. Two recordings can show similar peaks while sounding very different, because perceived loudness depends on duration, frequency balance, dynamics, and playback conditions. Do not use the picture to diagnose hearing, equipment health, clipping history, phase relationships, or mastering compliance.
For multichannel audio, the image is a combined peak overview rather than separate channel lanes. A bucket can take its minimum from one channel and its maximum from another, so the original audio is never visually split into discrete left and right strips. If you need separate channel waveforms, export each channel as its own track in an audio editor first.
Everything happens locally. The selected audio file is decoded with the Web Audio API AudioBuffer in your current browser tab, the peak extraction and SVG generation happen locally, and the download is plain text SVG. Nothing is sent to a server. Selecting a replacement file, changing an option, regenerating, or closing the page revokes the obsolete object URL so an older image cannot appear to represent new inputs. The tool refuses unsupported, corrupt, empty, or disguised files, and over-limit audio is rejected as a whole rather than silently truncated to the first few seconds.
If you're weighing options, Common Audio Waveform Generation Mistakes and Fixes covers this in detail.
If you're weighing options, What Can Go Wrong When Generating an Audio Waveform? covers this in detail.