Audio file formats and quality

Ask what the next person or program actually requires: file type, encoding, channel arrangement and any sample-rate or bit-depth restriction. Keep that request alongside the original filename and recording source. Avoid changing a file merely because another format sounds more professional.

There are three useful roles:

• Original: the file you received or obtained from the recording system, kept unchanged.
• Working copy: a version the editing software can process without repeatedly creating new lossy generations.
• Playback copy: a version chosen for a particular player, transfer limit or recipient.

One file may serve more than one role, but a converted derivative should not replace the retained original. Apple’s conversion documentation explicitly describes making another version while keeping the original. A suitable copy is a workflow choice, not a claim that the recording is authentic or complete.

A container organizes the file; an encoding describes how the audio data is represented. A codec is a program, algorithm or device that encodes or decodes the audio data. The Library of Congress identifies WAVE as a wrapper that can contain linear PCM and several compressed encodings. A .wav filename therefore does not prove that the audio is uncompressed, high resolution or free from earlier processing.

Where the software exposes them, inspect the actual encoding, sample rate, bit depth, channels and duration. Microsoft’s WAVEFORMATEX documentation treats these as separate fields. Some compressed encodings do not have a meaningful PCM-style bit-depth value, so an empty field is not permission to guess.

For an ordinary PCM working copy, WAV is a common option. FLAC is useful when the receiving software supports lossless compressed audio. Compatibility still depends on the actual file and application; an extension alone cannot settle it.

Lossless coding reduces storage while allowing the encoded audio samples to be reconstructed. Xiph.Org describes FLAC this way. Its compression level changes how the encoder searches for a compact representation; it is not a dial that progressively removes speech detail. There is no single percentage reduction that applies to every recording.

Lossy coding, including MP3 and AAC, discards information to reduce data. Audacity’s digital-audio documentation explains that MP3 uses a model of hearing to decide what information to omit. That does not mean every loss is harmless in every recording or for every later task.

If the only available source is an MP3, retain it. Decoding it into PCM WAV may be appropriate for editing, but the WAV represents the audio that can be decoded from that MP3. It does not recover an earlier, uncompressed recording. Sending that WAV through another lossy export can introduce another generation of loss; Apple warns that converting between compressed formats may reduce sound quality.

Return to a better original export when one exists. Otherwise, assess the recording you actually have rather than treating conversion as restoration.

Sample rate is the number of audio samples per second, usually shown in Hz or kHz. Audacity’s documentation distinguishes this from the audio frequencies those samples can represent. A file at 48 kHz contains 48,000 samples per second in each channel; it does not contain a 48 kHz audible tone.

For existing audio, start with the source rate unless the next task requires a change. Resampling calculates samples at a new rate. Changing how a rate is interpreted without proper conversion can affect playback speed and pitch. Increasing the export rate does not reconstruct source information that was never captured or has already been removed.

Bit depth describes sample precision in PCM audio. More available precision can be useful during processing, but choosing a larger output depth does not undo microphone noise, clipping or earlier lossy coding. Distinguish the editor’s internal processing format from the format needed for delivery.

Bitrate is a data rate, commonly expressed in kilobits per second, or kbps. It is different from bit depth. For uncompressed PCM, approximate audio-payload size is:

sample rate × bits per sample × channels × seconds ÷ 8

For a calculated 20-minute example, 48 kHz, 16-bit mono PCM contains 115.2 MB of audio payload; stereo doubles that to 230.4 MB. These decimal MB values exclude the container header and metadata. They describe storage, not how intelligible the recording is. A larger converted file can contain the same unresolved words.

A compressor effect changes the relationship between louder and quieter sound. Audacity’s Compressor documentation describes reducing dynamic range. That is a processing decision and can alter the recording’s sound.

A file encoder changes how audio is stored. Saving lossless compressed audio is not the same operation as applying a dynamic-range compressor. If someone says “compress the audio,” clarify whether they mean a smaller file or more even levels before changing anything.

Dither is different again. Audacity describes it as intentional low-level noise used to randomize quantization errors when reducing bit depth. It is not a method for removing hiss or improving a poor source by adding resolution. Export decisions depend on the processing and destination; repeatedly reducing bit depth or adding noise without understanding the settings is not a preservation strategy.

Two files can decode to the same audio samples while differing in container structure or metadata. Xiph.Org explicitly warns that FLAC does not preserve all non-audio WAVE information by default. Some workflows can retain additional metadata, but that requires deliberate support and verification.

Keep the original file even if the converted copy sounds identical. Preserve relevant recording-system exports and surrounding records separately when needed. A successful playback check answers whether the copy can be played; it does not establish that every original metadata field survived, or prove the recording’s origin or editing history.

This distinction matters when handing a file to an examiner. Explain which file is the original supplied source, which is the derivative and what conversion you performed.

Use a reversible sequence:

1. Keep the original in a separate location and give the derivative a distinct name.
2. Confirm what the receiving application needs. If no conversion is necessary, do not add one just to change the extension.
3. Export the required encoding, preserving the source channel arrangement and rate where the task permits. Avoid an accidental mono mix or a lossy default.
4. Reopen the exported file. Check duration, channel count and the reported encoding and settings against the intended result.
5. Listen at the beginning, middle, end and any critical timestamps. Check speech and timing, not just whether the file opens.
6. Keep a short note of the software and settings used, alongside the retained original.

As a specific example, Apple’s Music guide describes choosing an import encoding under Music → Settings → Files → Import Settings, then using File → Convert to create the selected version. The original and converted versions remain in the library. This is a music-library example, not a universal procedure for proprietary recorder exports or evidence handling.

If a conversion changes duration, playback speed or channel content unexpectedly, return to the original and investigate the settings before processing further.

Format conversion may solve compatibility while leaving noise, distortion or unclear speech unchanged. If that happens, compare the copy with the original and identify the exact passage that matters. Avoid repeatedly exporting different formats in the hope that the next extension will reveal the words.

For a file review, explain the recording source, original filename, important timestamps, changes already made and any deadline. Keep the original available. Media Medic’s review process has an audio engineer listen before invoicing; cleanup is paid work. The recording determines what improvement may be possible.

This guide compares official format documentation from Xiph.Org, the Library of Congress, Microsoft, Audacity and Apple. It separates encoding, numerical settings and file preservation, and calculates the PCM storage example from sample rate, bit depth, channels and duration.

It does not report a listening test, a restoration experiment or a forensic examination. The size example excludes headers and metadata. Application support and export behavior must be checked for the actual software and source; a general format guide cannot establish authenticity or promise recoverable speech.

Preserve the original, choose a copy that fits the next task, and verify what changed. Lossless coding can protect the audio presented to it. It cannot turn a damaged or previously reduced source into an earlier recording that you no longer have.