Radio & Phone Voice Effect

How Radio and Telephone Audio Filters Work: Bandpass, Distortion, and Bit Reduction

The human voice on a telephone call sounds nothing like the same voice recorded in a studio. The difference comes from deliberate bandwidth restriction. A standard telephone channel transmits audio between 300 Hz and 3400 Hz, as defined by the ITU-T P.342 recommendation for narrowband telephony. Everything below 300 Hz (the fundamental frequency of most male voices) and everything above 3400 Hz (the harmonics that give speech its clarity and air) is removed before transmission. AM radio occupies a similarly narrow band, historically limited to roughly 300 Hz to 3000 Hz on consumer receivers, though Ofcom announced in 2024 that UK AM stations can now apply for wider audio bandwidth up to 9 kHz. This tool replicates both bandwidth restrictions using a highpass and lowpass BiquadFilterNode pair, then layers distortion, static noise, and bit-depth reduction on top.

The Bandpass Filter: Why 300 to 3400 Hz Matters

A bandpass filter is the combination of a highpass filter (which removes low frequencies) and a lowpass filter (which removes high frequencies). The Web Audio API provides both as BiquadFilterNode types, and this tool chains them in series. The highpass filter removes everything below the low cutoff, and the lowpass removes everything above the high cutoff. The remaining band of frequencies is what passes through.

The telephone bandwidth of 300 to 3400 Hz was not chosen at random. The ITU-T G.711 PCM standard, which has defined digital telephony since 1988, specifies a sampling rate of 8000 samples per second with 8-bit quantization. By the Nyquist theorem, an 8000 Hz sample rate can represent frequencies up to 4000 Hz, but the actual usable bandwidth is reduced to 3400 Hz by the anti-aliasing filters in the codec. The 300 Hz lower boundary exists because frequencies below 300 Hz in speech carry very little intelligibility information and consume valuable transmission power. The result is a 3100 Hz wide channel that transmits speech with acceptable clarity but strips away the fullness, warmth, and presence that a studio microphone captures.

AM radio occupies a similar but slightly wider band. The NRSC AM Bandwidth Study found that the majority of consumer AM receivers in the United States have audio bandwidths of 5 kHz or less, with many receivers producing usable output only up to 2.5 to 3 kHz. This is why AM radio has that characteristic tinny, compressed sound. The tool defaults to a 300 to 3000 Hz bandpass for the Radio preset, which matches the typical consumer AM receiver response curve.

Distortion and Compression via WaveShaper

Bandpass filtering alone produces a thin, nasal sound, but it does not produce the compressed, slightly distorted character of a real radio or phone signal. Telephone codecs apply companding (compression and expanding) using A-law or mu-law algorithms defined in G.711. These algorithms quantize quiet signals with more precision than loud signals, which creates a subtle but audible nonlinear distortion on the decoded audio. AM radio transmitters use audio processing chains with compression and limiting to maximize modulation depth within the allowed bandwidth, which adds harmonic distortion and reduces dynamic range.

This tool simulates both effects using a WaveShaperNode with a tanh-based soft clipping curve. The Drive slider controls the amount of distortion applied. At low values (20 to 30%), the effect is a gentle compression that reduces dynamic range without obvious clipping. At higher values (50 to 80%), the distortion becomes audible, producing the gritty, overdriven character of a cheap walkie-talkie or an overmodulated AM transmitter. The WaveShaper uses 2x oversampling to reduce aliasing artifacts that would otherwise be introduced by the nonlinear processing.

Static Noise and Signal Degradation

Real radio transmissions pick up atmospheric noise, electrical interference, and signal degradation from weak reception. Telephone lines introduce line noise and crosstalk. The Static Noise slider mixes white noise into the processed signal at a user-controlled level. At 4% (the Radio preset default), the noise is barely perceptible but adds a subtle texture that breaks the silence between words. At 12% (the Old Radio preset), the static is clearly audible and creates the sense of a distant, fading signal. The noise is generated sample-by-sample using a uniform random distribution, which produces a broadband hiss similar to thermal noise in an analog receiver.

Bit-Depth Reduction and Sample-Rate Reduction

The telephone codec character comes from more than just bandwidth restriction. G.711 quantizes each sample to 8 bits using logarithmic companding, which produces 256 discrete amplitude levels. A 16-bit studio recording has 65,536 levels. When you reduce the bit depth from 16 to 8, the quantization error increases and becomes audible as a gritty, grainy texture on quiet passages. The Bit Depth slider lets you reduce the effective quantization from 16-bit down to 4-bit (only 16 levels), which produces extreme quantization noise that sounds like a broken digital connection.

Sample-rate reduction (also called decimation) simulates the low sample rate of telephone codecs and digital radio systems. G.711 operates at 8000 Hz. If your source file is 44100 Hz or 48000 Hz, reducing the effective sample rate to 8000 Hz means the codec can only represent frequencies up to 4000 Hz. But the tool achieves this by holding each sample for multiple consecutive output samples (a sample-and-hold process), which creates aliasing artifacts that mirror high frequencies back into the audible band. This aliasing is what gives low-bandwidth digital audio its metallic, robotic quality. The Sample Rate Reduction slider lets you control how aggressively the sample rate is reduced, from 0% (full quality) to 75% (quarter sample rate).

Understanding the Presets

The four presets map to specific real-world audio transmission scenarios. Each one sets all six parameters to values that match the characteristics of that transmission type.

• Radio uses a 300 to 3000 Hz bandpass with mild distortion (30%) and light static (4%). No bit reduction or sample-rate reduction is applied. This produces the sound of a clean AM radio signal received on a decent consumer radio. Use this for radio broadcast voiceovers, news report simulations, or any scene where the voice needs to sound like it is coming through a car radio or kitchen radio.
• Phone uses a 300 to 3400 Hz bandpass with heavier distortion (55%), moderate static (6%), 8-bit quantization, and 50% sample-rate reduction. This matches the G.711 codec characteristics and produces the classic telephone call sound. Use this for phone conversation scenes, voicemail recordings, or podcast segments that simulate a remote interview over a phone line.
• Old Radio uses a narrower 400 to 2500 Hz bandpass with warm distortion (65%), prominent static (12%), 12-bit quantization, and 25% sample-rate reduction. This simulates a vintage tube radio from the 1940s or 1950s with a worn speaker and marginal reception. Use this for period pieces, historical dramas, or lo-fi music production where you want a distinctly aged, degraded radio sound.
• Walkie-Talkie uses a very narrow 500 to 2200 Hz bandpass with aggressive distortion (80%), moderate static (8%), 6-bit quantization, and 50% sample-rate reduction. This simulates a cheap CB radio or consumer walkie-talkie with a tiny speaker and heavy compression. Use this for two-way radio scenes, military communication simulations, or any context where the voice needs to sound like it is coming through a low-quality handheld transceiver.

After selecting a preset, you can fine-tune any individual parameter without losing the starting point. The preset label stays active until you manually adjust a slider, at which point it clears to indicate you are in custom territory. If you need to pitch-shift the processed result before exporting, use the Pitch Shifter tool after downloading the WAV file from this tool.

Rendering and Exporting the Processed Audio

Clicking "Apply Effect" runs the bandpass filter and WaveShaper distortion through an OfflineAudioContext, which renders the entire file in memory rather than playing it in real time. This guarantees a consistent output regardless of your CPU load during rendering. After the offline render completes, the bit-depth reduction, sample-rate reduction, and static noise are applied as post-processing steps on the rendered buffer. This two-stage approach gives the filters and distortion the full resolution of the Web Audio graph, while the degradation effects are applied on the final output where they produce the most realistic artifacts.

The processed buffer is available for A/B preview using the Original and Effect toggle buttons. The download button exports the processed audio as a 16-bit PCM WAV file. The output duration matches the input duration exactly. If your source file is 2:30 long, the processed output is also 2:30 long.

For additional processing before or after applying the radio or phone effect, the Vinyl & Tape Warble Simulator adds pitch modulation and tape saturation for a more complete vintage degradation chain. The Audio Trimmer can cut the processed file to the exact length you need for a film scene or podcast segment. Both tools accept WAV files and run entirely in-browser, so you can chain multiple processing steps without any file uploads.

Real-World Applications for Music and Audio Production

Radio and phone effects are used across music production, film audio, podcasting, and game audio design. In music, the telephone voice effect is a staple of hip-hop and pop production, where a verse or intro section is processed to sound like a phone call before the full-bandwidth beat drops. Producers like Metro Boomin and Dr. Luke have used this technique on charting singles to create contrast between sections. The effect works because the listener's brain immediately recognizes the restricted bandwidth as a telephone signal, which creates a sense of distance and intimacy that the full-bandwidth chorus then shatters.

In film and television, radio and phone effects are applied to dialogue to match the visual context of the scene. A character speaking on a walkie-talkie needs audio that matches the visual prop, and a news reporter on the radio needs that AM broadcast quality. Sound designers typically use dedicated plugins like iZotope RX or Accusonus ERA for these tasks, but this tool provides the same core processing chain (bandpass filtering, distortion, noise, and bit reduction) for free in the browser.

For podcasters, the phone effect is useful for simulating remote interviews when the actual recording quality is too clean. Applying a 300 to 3400 Hz bandpass with mild distortion and a small amount of static makes a studio-quality recording sound like it was captured over a phone line, which can help maintain consistency when a podcast episode mixes studio recordings with actual phone calls. Before exporting the processed file for a podcast, check its loudness with the LUFS Loudness Meter to ensure it meets your platform target. The distortion and noise added by this tool can push the integrated loudness up by 1 to 3 LUFS depending on the settings.