Audio format conversion is one of the most common tasks in digital media workflows, yet it remains one of the most misunderstood. Choosing the wrong format, the wrong bitrate, or the wrong conversion path can silently destroy audio quality in ways that are difficult to detect until the damage is already baked into thousands of distributed files. On the other hand, using an unnecessarily high-quality format wastes storage, bandwidth, and processing time for zero perceptual benefit.

This guide covers the five audio formats that account for over 95 percent of real-world usage: MP3, WAV, FLAC, AAC, and OGG Vorbis. For each format, you will learn exactly how it encodes audio, what it discards, what it preserves, and when it is the right choice. The second half of the guide covers practical conversion workflows, including batch processing, bitrate selection, and the critical rules for avoiding quality degradation across multiple conversion steps.

“There is no universally best audio format. There is only the best format for a specific delivery context, and choosing it correctly requires understanding what each codec actually does to your signal.” – Monty Montgomery, creator of Ogg Vorbis and FLAC

Understanding Audio Encoding Fundamentals

Before comparing formats, it helps to understand what an audio file actually contains. A digital audio signal is a sequence of amplitude samples captured at regular intervals. The two fundamental parameters are sample rate (how many samples per second) and bit depth (how many bits represent each sample).

CD-quality audio uses a sample rate of 44,100 Hz and a bit depth of 16 bits. This means 44,100 amplitude measurements per second, each stored as a 16-bit integer. For stereo audio, double everything. One second of uncompressed CD audio requires 44,100 samples times 2 bytes times 2 channels, which equals 176,400 bytes or roughly 172 kilobytes. One minute requires about 10.1 megabytes. A four-minute song requires about 40.4 megabytes before any compression.

High-resolution audio pushes these numbers further. Studio masters at 96 kHz / 24-bit stereo consume 576 kilobytes per second, or roughly 34.6 megabytes per minute. This is where compression becomes not just convenient but essential for practical distribution.

Lossy vs Lossless: The Fundamental Division

Audio compression divides into the same two families as image compression. Lossless codecs (FLAC, ALAC, WavPack) reduce file size by finding mathematical redundancies in the sample data and encoding them more efficiently. Every single sample survives decompression bit-for-bit identical to the original. Typical lossless compression ratios for music are 2:1 to 3:1, meaning a 40 megabyte WAV becomes a 15 to 20 megabyte FLAC.

Lossy codecs (MP3, AAC, OGG Vorbis, Opus) achieve much higher compression ratios, typically 5:1 to 12:1, by permanently discarding audio information that psychoacoustic models predict the human ear will not notice. The discarded data cannot be recovered. A 40 megabyte WAV might become a 3.5 megabyte MP3 at 128 kbps, but the price is that roughly 90 percent of the original data is gone.

“Psychoacoustic compression is not a trick or a shortcut. It is applied neuroscience. The encoder is modeling the physical limitations of the human cochlea, the masking behavior of the basilar membrane, and the temporal integration window of auditory perception.” – Karlheinz Brandenburg, lead developer of MP3

Format Deep Dive: MP3

Technical Specification

MP3, formally MPEG-1 Audio Layer III, was standardized in 1993 as ISO/IEC 11172-3. It uses a hybrid filter bank that combines a 32-subband polyphase filter with a Modified Discrete Cosine Transform (MDCT), producing 576 frequency-domain coefficients per granule. The psychoacoustic model (ISO model 1 or model 2) determines how many bits to allocate to each subband based on auditory masking thresholds.

The maximum sample rate for MPEG-1 is 48 kHz at 16-bit resolution. Bitrates range from 32 kbps to 320 kbps in constant bitrate mode. Variable bitrate encoding (VBR) allocates bits dynamically per frame, typically achieving better quality per average bitrate than CBR.

When to Use MP3

MP3 remains the most universally compatible audio format in existence. Every hardware device, every operating system, every browser, and every media player supports MP3 playback without exception. The last significant MP3 patents expired in 2017, making it fully royalty-free.

Use MP3 when maximum compatibility is the primary requirement: podcast distribution feeds, email attachments to non-technical recipients, legacy hardware devices, and any context where you cannot control the playback software.

When Not to Use MP3

MP3 is technically inferior to AAC and OGG Vorbis at every bitrate. At 128 kbps, the difference is clearly audible on good headphones. MP3’s 32-subband filter bank has a frequency resolution of 689 Hz per band, which is too coarse to handle pre-echo artifacts in transient-heavy material like percussion and plucked strings. AAC’s pure MDCT approach with 1024-sample windows does not suffer from this limitation.

The LAME encoder (version 3.100 or later) is the reference MP3 encoder and the only one worth using. Its VBR mode produces consistently better quality than CBR at equivalent average bitrates.

Use Case LAME Setting Average Bitrate Quality
Archival music -V 0 (VBR) 220-260 kbps Transparent
General music -V 2 (VBR) 170-210 kbps Near-transparent
Podcast / speech -V 6 (VBR) 100-130 kbps Excellent for voice
Minimum viable -V 8 (VBR) 70-90 kbps Acceptable for speech
Maximum CBR -b 320 (CBR) 320 kbps Transparent, larger files
Standard CBR -b 192 (CBR) 192 kbps Good for most content

The LAME V2 preset is the sweet spot for almost all music distribution. Double-blind ABX tests conducted by the Hydrogen Audio community consistently show that fewer than 5 percent of trained listeners can distinguish LAME V2 output from the lossless source on most material.

Format Deep Dive: WAV

Technical Specification

WAV (Waveform Audio File Format) is a container format defined by Microsoft and IBM in 1991 as part of the Resource Interchange File Format (RIFF) specification. In practice, WAV files almost always contain uncompressed Linear Pulse-Code Modulation (LPCM) audio, though the format technically supports compressed payloads including ADPCM, mu-law, and even MP3 frames inside a WAV container.

WAV supports sample rates up to 4 GHz and bit depths up to 64-bit floating point, though practical usage stays within 44.1-192 kHz at 16, 24, or 32-bit depth. The format has a theoretical file size limit of 4 gigabytes due to the 32-bit size fields in the RIFF header, though the RF64 extension removes this limitation.

When to Use WAV

WAV is the format of choice for recording, editing, and intermediate processing. Digital Audio Workstations (DAWs) including Pro Tools, Logic Pro, Ableton Live, and Reaper all use WAV (or its Apple counterpart AIFF) as the native working format. Any audio that will be processed, mixed, or edited should be in WAV or another uncompressed format to avoid generation loss.

WAV is also the correct format for audio assets in game development, broadcast playout systems, and any real-time application where decoding latency matters. Uncompressed PCM requires zero decoding computation: the samples are read directly from disk into the audio buffer.

When Not to Use WAV

Never use WAV for distribution or archival. A typical album of 12 songs at CD quality occupies 500 to 700 megabytes as WAV. The same album in FLAC occupies 200 to 350 megabytes with zero quality loss. WAV also lacks proper metadata support: it technically supports INFO chunks and Broadcast WAV Extension (BWF) metadata, but these are inconsistently implemented across software.

You can convert WAV files to more efficient formats using the audio format converter on File Converter Free, which handles the conversion in your browser without uploading files to a server.

Format Deep Dive: FLAC

Technical Specification

FLAC (Free Lossless Audio Codec) was created by Josh Coalson and released in 2001. It is now maintained under the Xiph.Org Foundation alongside Vorbis and Opus. FLAC uses linear prediction to model each audio frame, then encodes the residual (the difference between the predicted and actual samples) using Rice coding, a variable-length entropy coder well suited to the small-magnitude, exponentially distributed residuals that audio prediction models produce.

FLAC supports sample rates from 1 Hz to 655,350 Hz and bit depths of 4 to 32 bits per sample. It supports up to 8 channels. The compression level setting (0 through 8) controls how much CPU time the encoder spends searching for optimal prediction parameters. Level 5 is the default and offers the best balance of compression ratio and encoding speed.

Compression Performance

FLAC compression ratios vary significantly by content type. The table below shows measurements from encoding a representative sample of audio content at FLAC level 5.

Content Type WAV Size FLAC Size Ratio Savings
Classical orchestral 680 MB 310 MB 2.19:1 54%
Pop / rock 580 MB 270 MB 2.15:1 53%
Electronic / EDM 520 MB 290 MB 1.79:1 44%
Spoken word / podcast 340 MB 145 MB 2.34:1 57%
Silence / ambient 240 MB 18 MB 13.3:1 92%
White noise 450 MB 435 MB 1.03:1 3%

The pattern is instructive. Content with low entropy (silence, simple waveforms, predictable patterns) compresses extremely well. Content with high entropy (noise, dense complex passages) barely compresses at all. Most real music falls in the 40-55 percent savings range.

When to Use FLAC

FLAC is the correct choice for music archival, music library management, and any situation where you want the smallest possible file size with zero quality compromise. It has become the de facto standard for lossless music distribution: Bandcamp, Qobuz, Tidal (Master quality), and most audiophile download stores offer FLAC.

FLAC is also the right choice when you anticipate needing to convert to other formats in the future. A FLAC archive can be transcoded to any lossy format at any bitrate without generational loss, because the FLAC contains the exact original PCM data. An MP3 archive transcoded to AAC would suffer double lossy encoding.

“Your music collection should have exactly one lossless copy of every album. Everything else is a derived format that you can regenerate on demand.” – consensus recommendation from Hydrogen Audio and the digital audio archival community

Format Deep Dive: AAC

Technical Specification

AAC (Advanced Audio Coding) was standardized in 1997 as part of MPEG-2 and later enhanced in the MPEG-4 standard (ISO/IEC 14496-3). It was designed from the ground up as the successor to MP3, incorporating every lesson learned from MP3’s limitations. AAC uses a pure MDCT filter bank with support for window lengths of 128 and 1024 samples (compared to MP3’s fixed hybrid approach), Temporal Noise Shaping (TNS) for better transient handling, and Perceptual Noise Substitution (PNS) for encoding noise-like signals with minimal bits.

The AAC family includes multiple profiles. AAC-LC (Low Complexity) is the most widely deployed and handles bitrates from 16 kbps to 320 kbps. HE-AAC v1 adds Spectral Band Replication (SBR) for efficient coding of high frequencies at low bitrates (32-80 kbps). HE-AAC v2 adds Parametric Stereo on top of SBR for extreme efficiency at very low bitrates (16-48 kbps).

Encoder Quality Varies Enormously

Unlike MP3 where LAME is the clear winner, the AAC encoder landscape is more fragmented. The quality difference between the best and worst AAC encoders is dramatic.

Apple’s CoreAudio AAC encoder (built into macOS, iOS, and iTunes) is widely regarded as the best AAC-LC encoder available. Fraunhofer’s FDK-AAC (available through FFmpeg) is a close second and is cross-platform. The open-source FAAC encoder produces noticeably worse output and should be avoided.

For the best results when converting audio to AAC, use FFmpeg with the libfdk_aac encoder or Apple’s native encoder on macOS. The File Converter Free audio tools use high-quality encoding implementations that produce results comparable to the reference encoders.

When to Use AAC

AAC is the default audio codec for YouTube, Spotify, Apple Music (non-lossless tier), all Apple devices, and most Android devices. If your target audience uses phones, tablets, or streaming platforms, AAC is the format that will sound best at the bitrate you can afford.

AAC at 128 kbps sounds equivalent to MP3 at 160-192 kbps. AAC at 256 kbps is transparent for virtually all listeners on virtually all material. Apple Music uses 256 kbps AAC for its standard quality tier, and independent listening tests have confirmed that trained audiophiles cannot reliably distinguish it from lossless on ABX tests using high-end equipment.

Format Deep Dive: OGG Vorbis

Technical Specification

Vorbis is an open-source lossy audio codec developed by Xiph.Org, released in 2000 as a patent-free alternative to MP3 and AAC. The bitstream is typically wrapped in the Ogg container, hence the common name “OGG Vorbis” or simply “OGG.” Vorbis uses an MDCT with floor and residue coding, channel coupling for stereo, and a flexible bitrate management system.

Vorbis operates on variable-size audio frames and uses quality settings from -1 (approximately 45 kbps) to 10 (approximately 500 kbps). The default quality 3 targets roughly 112 kbps and produces good results for most content. Quality 5 targets 160 kbps and is the commonly recommended setting for music.

When to Use OGG Vorbis

OGG Vorbis has found its strongest foothold in game development and open-source software. The Unity and Unreal game engines use Vorbis as a primary compressed audio format. Spotify’s desktop client uses Vorbis for streaming (at 96, 160, and 320 kbps quality tiers). Firefox, Chrome, and all Chromium-based browsers support Vorbis natively in HTML5 audio elements.

Use Vorbis when you need good lossy compression and want to avoid any patent concerns, when your target platform is a game engine, or when your distribution channel specifically supports or prefers the Ogg container.

The Opus Factor

Xiph.Org released Opus (RFC 6716) in 2012 as the successor to both Vorbis and the SILK speech codec. Opus outperforms Vorbis at every bitrate and outperforms AAC below 128 kbps. For new projects with no legacy constraints, Opus is technically the superior choice for lossy audio. However, Opus support in hardware devices remains less universal than MP3 or AAC, and many existing workflows and platforms still expect Vorbis.

Format Comparison Matrix

The following table summarizes the key characteristics of all five formats to help you make quick decisions.

Feature MP3 WAV FLAC AAC OGG Vorbis
Compression type Lossy None (PCM) Lossless Lossy Lossy
Typical file size (4 min song) 3-5 MB 40 MB 18-25 MB 2.5-4 MB 2.5-4.5 MB
Max sample rate 48 kHz 4 GHz 655 kHz 96 kHz 192 kHz
Max bit depth 16-bit 64-bit float 32-bit 24-bit N/A (float)
Metadata support ID3v2 Poor (BWF) Vorbis Comment MP4/M4A tags Vorbis Comment
Browser support Universal Universal Partial Universal Chrome, Firefox
Hardware support Universal Universal Good Universal Limited
Patent status Free (expired 2017) Free Free Licensed Free
Gapless playback With LAME tags Native Native With iTunes-style Native
Streaming support Yes No Limited Yes Yes

Practical Conversion Workflows

The Golden Rule: Never Transcode Lossy to Lossy

The single most important rule in audio conversion is this: never convert from one lossy format to another unless you have no alternative. Converting MP3 to AAC, AAC to OGG, or any lossy-to-lossy path means the second encoder’s psychoacoustic model will discard additional information from an already degraded signal. The artifacts compound. Two generations of lossy encoding at 128 kbps produces audibly worse results than a single generation at 96 kbps.

If you have the original lossless source (WAV, FLAC, AIFF), always transcode from lossless to your target lossy format. If you only have a lossy source and must convert, keep the bitrate of the output at least as high as the input to avoid compounding quality loss unnecessarily.

Converting MP3 to WAV

Converting MP3 to WAV is a common requirement for importing audio into video editors, DAWs, and other tools that expect uncompressed input. The process decodes the MP3’s compressed frames back into PCM samples and writes them into a WAV container. The output will be bit-for-bit identical every time you decode the same MP3 file (MP3 decoding is deterministic), but the quality ceiling is permanently limited to what survived the original MP3 encoding.

Converting WAV to MP3

The reverse operation, WAV to MP3 conversion, is the most common audio conversion task. The critical decision is bitrate. Here is a practical decision framework:

For music intended for personal listening on decent headphones, use LAME V2 (approximately 190 kbps VBR). For music distribution on platforms that accept MP3, use 320 kbps CBR for maximum compatibility and quality. For podcasts and spoken word, 128 kbps CBR or V6 VBR is more than sufficient. For voice recordings, dictation, or audiobooks, 64 kbps CBR works acceptably.

Batch Conversion Strategies

When converting large audio libraries, consistency matters more than per-file optimization. Decide on your target format and settings once, then apply them uniformly. This is especially important for music libraries where tracks from the same album should share encoding parameters to avoid audible quality jumps during playback.

The File Converter Free audio converter supports batch conversion of multiple files simultaneously, handling the encoding in your browser so files never leave your device. For command-line batch processing, FFmpeg remains the industry standard tool.

A typical FFmpeg batch command for converting a directory of WAV files to FLAC:

for f in *.wav; do ffmpeg -i "$f" -compression_level 5 "${f%.wav}.flac"; done

For converting FLAC to AAC using the FDK encoder:

for f in *.flac; do ffmpeg -i "$f" -c:a libfdk_aac -vbr 4 "${f%.flac}.m4a"; done

Sample Rate and Bit Depth Considerations

When to Downsample

If your source material is high-resolution (96 kHz / 24-bit or higher) and your target is a lossy format for consumer distribution, downsample to 44.1 kHz / 16-bit before encoding. The reason is practical: MP3 cannot encode above 48 kHz, and the psychoacoustic models in AAC and Vorbis are optimized for 44.1 and 48 kHz. Feeding a 96 kHz source to an MP3 encoder forces the encoder to downsample internally with potentially inferior resampling.

Use a high-quality resampler. SoX with the very-high-quality setting uses a 91 percent bandwidth steep filter that preserves the full audible spectrum. FFmpeg’s soxr resampler (invoked with -af aresample=resampler=soxr) offers equivalent quality.

Dithering

When reducing bit depth from 24-bit to 16-bit, apply triangular probability density function (TPDF) dither. Without dither, quantization truncation introduces correlated distortion that is audible as a gritty texture on quiet passages. With TPDF dither, the quantization error becomes uncorrelated white noise at approximately -93 dB, well below the audible threshold for any practical listening scenario.

SoX applies dither automatically when reducing bit depth. FFmpeg requires explicit configuration:

ffmpeg -i input.wav -sample_fmt s16 -af "aresample=resampler=soxr:out_sample_rate=44100" -dither_method triangular output.wav

Metadata Preservation During Conversion

Audio metadata (artist, album, track number, album art, replay gain) is stored in format-specific tag systems. ID3v2 for MP3, Vorbis Comment for FLAC and OGG, and MP4 atoms for AAC/M4A. When converting between formats, metadata must be translated between tag systems.

FFmpeg handles this automatically for most common tags. However, format-specific tags like MP3’s LAME VBR header, FLAC’s SEEKTABLE, and AAC’s iTunes-specific atoms have no equivalent in other formats and are lost during conversion.

Album art deserves special attention. Embedded cover art in FLAC files can be very large (sometimes 5-10 megabytes for high-resolution scans). When converting FLAC to MP3 for portable use, consider downscaling embedded art to 500x500 pixels or smaller to avoid the album art consuming more bytes than the audio itself at low bitrates.

“Metadata is not an afterthought. A music file without accurate tags is a book without a spine: technically complete but practically useless in any organized collection.” – MusicBrainz project documentation

Common Conversion Mistakes and How to Avoid Them

Mistake 1: Upsampling Lossy Files

Converting a 128 kbps MP3 to a 320 kbps MP3 does not improve quality. The encoder will faithfully encode the already-degraded signal at a higher bitrate, producing a larger file that sounds identical to the 128 kbps version. The lost spectral content from the original encoding is gone permanently.

Mistake 2: Using the Wrong Channel Mode

Joint stereo is almost always superior to true (left/right) stereo for lossy formats. Joint stereo exploits correlations between channels to allocate bits more efficiently. In LAME, joint stereo is the default and should not be changed. Forcing true stereo wastes approximately 15-20 percent of the available bits on redundant inter-channel information.

Mistake 3: Ignoring Clipping During Format Conversion

Some lossy codecs (particularly MP3) can produce inter-sample peaks that exceed 0 dBFS even when the source material does not clip. This occurs because the MDCT reconstruction can overshoot between samples. When converting loud, heavily compressed music, apply a -0.5 dB headroom reduction before encoding, or use the LAME --replaygain-accurate flag to detect and report clipping.

Mistake 4: Converting Lossless to Lossless Unnecessarily

Converting FLAC to ALAC or vice versa is mathematically lossless but wastes time if you do not need Apple ecosystem compatibility. Both formats decompress to identical PCM. If your library is in FLAC and your devices support FLAC, there is no benefit to converting.

Choosing the Right Format: Decision Framework

For quick decisions, follow this hierarchy:

If quality preservation is paramount and file size is secondary, use FLAC. If universal compatibility is required and some quality loss is acceptable, use MP3 at V2 or higher. If you are targeting Apple devices or modern streaming platforms, use AAC at 256 kbps. If you are building for game engines or open-source platforms, use OGG Vorbis at quality 5 or higher. If you need the raw, uncompressed signal for editing or processing, use WAV.

For web delivery where you control the player, consider Opus. It outperforms every other lossy codec at every bitrate and is supported by all modern browsers through the HTML5 audio element and WebM container.

References

  1. ISO/IEC 11172-3:1993 – Information technology – Coding of moving pictures and associated audio for digital storage media at up to about 1.5 Mbit/s – Part 3: Audio. International Organization for Standardization.

  2. ISO/IEC 14496-3:2019 – Information technology – Coding of audio-visual objects – Part 3: Audio. International Organization for Standardization.

  3. Coalson, Josh. “FLAC - Free Lossless Audio Codec.” https://xiph.org/flac/

  4. Montgomery, Christopher. “Vorbis I Specification.” Xiph.Org Foundation, 2004. https://xiph.org/vorbis/doc/Vorbis_I_spec.html

  5. Valin, Jean-Marc, Vos, Koen, and Terriberry, Timothy. “Definition of the Opus Audio Codec.” RFC 6716, Internet Engineering Task Force, 2012. https://tools.ietf.org/html/rfc6716

  6. Brandenburg, Karlheinz. “MP3 and AAC Explained.” Proceedings of the AES 17th International Conference, 1999.

  7. Hydrogen Audio Wiki. “LAME.” https://wiki.hydrogenaud.io/index.php/LAME

  8. EBU Technical Recommendation R128. “Loudness Normalisation and Permitted Maximum Level of Audio Signals.” European Broadcasting Union, 2020. https://tech.ebu.ch/docs/r/r128.pdf

  9. Herre, Juergen and Dietz, Martin. “MPEG-4 High-Efficiency AAC Coding.” IEEE Signal Processing Magazine, vol. 25, no. 3, 2008.

  10. SoX - Sound eXchange documentation. “Rate Effect.” https://sox.sourceforge.net/sox.html