What is the RIFF/WAVE format?

WAV is a RIFF (Resource Interchange File Format) container. It begins with the ASCII tag RIFF, a 4-byte file size, then WAVE, followed by sub-chunks. The fmt chunk holds the audio parameters and the data chunk holds the raw samples.

How is the duration calculated?

Duration in seconds equals the data chunk size in bytes divided by the byte rate, where byte rate is sampleRate × channels × (bitDepth / 8). For example a 44100 Hz, 16-bit stereo file has a byte rate of 176400 bytes per second.

What does the audio format number mean?

The format code in the fmt chunk identifies the encoding: 1 is uncompressed PCM, 3 is IEEE float, 6 is A-law, 7 is mu-law, and 65534 is WAVE_FORMAT_EXTENSIBLE. The inspector labels the common codes for you.

Why might the fmt chunk not be right after WAVE?

The WAV spec allows other chunks (like LIST or fact) before fmt or data. The inspector walks the chunk list by reading each chunk ID and size, so it finds fmt and data wherever they appear.

Is the audio uploaded to a server?

No. The header is parsed entirely in your browser with the FileReader and DataView APIs. No bytes are sent anywhere, so the tool works offline and keeps your audio private.

What is the WAV Header Inspector?

Free WAV header inspector that decodes the RIFF/WAVE header of a .wav file in your browser — showing audio format, sample rate, channel count, bit depth, byte rate, block align and exact duration. Runs locally with DataView, no upload. It runs free in your browser on Gera Tools, with nothing uploaded.

WAV Header Inspector

Name: WAV Header Inspector
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Decode the technical header of a WAV file — its sample rate, channel count, bit depth, encoding and exact duration — without any player or upload. The inspector reads the RIFF/WAVE structure directly with the browser’s DataView API and stays entirely on your device.

How it works

A WAV file is a RIFF container. The first 12 bytes are the ASCII tag RIFF, a little-endian 4-byte total size, and the ASCII tag WAVE. After that comes a sequence of chunks, each one a 4-character ID, a little-endian 4-byte size, and that many bytes of payload (padded to an even length).

The two chunks that matter are:

fmt   → audio parameters
  audioFormat  (2 bytes)  1=PCM, 3=float, 6=A-law, 7=mu-law, 65534=extensible
  numChannels  (2 bytes)
  sampleRate   (4 bytes)
  byteRate     (4 bytes)  = sampleRate × channels × bitDepth/8
  blockAlign   (2 bytes)  = channels × bitDepth/8
  bitsPerSample(2 bytes)

data  → the raw samples; its size gives the audio payload length

The inspector walks the chunk list rather than assuming a fixed layout, so it correctly handles files that place LIST or fact chunks before fmt or data.

Duration

Playback length comes straight from the data chunk:

duration (s) = data chunk size (bytes) ÷ byteRate

Because byteRate = sampleRate × channels × bitDepth / 8, a 10-second 44.1 kHz 16-bit stereo clip has a data chunk of 10 × 44100 × 2 × 2 = 1,764,000 bytes.

Reading each field and what it means

sampleRate

The number of audio samples per second, per channel. Common values:

Sample rate	Typical use
8,000 Hz	Telephony, voice over IP
22,050 Hz	Low-quality audio, older game audio
44,100 Hz	CD audio, most consumer audio
48,000 Hz	Professional video and broadcast standard
96,000 Hz	High-resolution audio production

If you see an unexpected sample rate like 22,050 Hz on a file you expected to be broadcast-quality, the file was either recorded at that rate or was downsampled. Resampling to a higher rate does not recover detail that was never captured.

numChannels

1 = Mono
2 = Stereo
6 = 5.1 surround (common in film audio)
8 = 7.1 surround

For surround formats, channel ordering follows the standard WAVE speaker assignment table (left front, right front, center, low-frequency, left back, right back, etc.).

bitsPerSample

8-bit: unsigned integer, low dynamic range, rarely used in modern audio
16-bit: CD standard, signed integer, 96 dB dynamic range
24-bit: studio production standard, signed integer, about 144 dB dynamic range
32-bit: often IEEE float (audioFormat = 3 rather than 1)

A 32-bit PCM file with audioFormat = 1 is signed integer with enormous headroom; a 32-bit file with audioFormat = 3 is floating-point, which can represent values outside the -1 to +1 range without clipping (used in intermediate processing stages).

blockAlign

Bytes per sample frame across all channels: numChannels × bitsPerSample / 8. For 16-bit stereo, blockAlign is 4 bytes. Audio players use this to seek to a specific time position without scanning through the data chunk.

audioFormat codes

1 (PCM): Uncompressed linear PCM. The most common and universally supported format.
3 (IEEE float): 32-bit or 64-bit floating-point samples. Common in digital audio workstations for headroom during mixing.
6 (A-law) / 7 (mu-law): Logarithmic compression used in telephony. Effectively 8-bit samples with companded dynamic range.
65534 (WAVE_FORMAT_EXTENSIBLE): A wrapper that stores the real format as a sub-format GUID in the extended fmt chunk. Used for high-channel-count and high-bit-depth audio where the original format codes are insufficient.

Tips

A mismatch between the stored byteRate and sampleRate × channels × bitDepth/8 usually means a malformed or non-standard encoder — the inspector shows both so you can compare.
Format code 65534 (extensible) wraps a real sub-format GUID; the bit depth and channels are still valid in the fmt chunk.
Very large declared sizes that exceed the actual file length indicate truncation or a streaming WAV written before the final size was known (a common output of live recording or network streaming that stopped unexpectedly).
LIST chunks carry metadata such as artist, copyright, and encoding software in plain ASCII; fact chunks appear in compressed WAV formats and carry the sample count. Neither is required for PCM files, but they appear in practice and the inspector skips them correctly to find fmt and data.