Audio formats and limits

Audivane imports a local audio file, decodes it once, and runs both codec chains from that single decoded source. This page lists exactly what it accepts and what it refuses. Nothing is uploaded or fetched remotely — playlists and remote media references are not accepted on any platform.

Accepted formats

The browser and desktop workbench accept WAV, FLAC, MP3, M4A/MP4 audio, ADTS AAC and Ogg/Opus (.ogg, .oga, .opus).

The browser identifies files by container signature rather than by file name or reported MIME type, so a renamed file is still handled correctly. The desktop app selects an explicit demuxer from the extension and requires the probe facts to match.

Size and duration ceilings

Boundary	Limit
Original local file	512 MiB
Track duration	up to 30 minutes (1800 seconds)
Final delivered audio, including accumulated codec padding	1800.5 seconds

Files beyond the envelope are refused, not served as shortened excerpts. Decoders carry deliberate time and byte guards before any audio is accepted, so an overlong file can never quietly become an accepted prefix. If a decoder produces slightly more audio than a container advertised — which can happen with AAC — the accepted samples remain in the prepared comparison instead of being silently chopped.

The largest layouts are memory-hungry: a full 30-minute, 96 kHz stereo delivery is about 691 MB per side before playback buffers. Shorter sources, an explicit 48 kHz delivery and mono material reduce the working set. These limits are policy ceilings, not a guarantee that every host or browser tab can allocate the largest layout; if the browser runs out of memory, the import fails with actionable guidance rather than reporting shortened success.

Sample rate and channels

• Mono/stereo layout is preserved. Mono sources are not padded to stereo, and stereo sources are not collapsed.
• Native-rate mode permits 44.1, 48 and 96 kHz. Other source rates need an explicit 48 kHz conversion, which you choose — nothing is resampled silently.
• Multichannel sources are refused rather than silently downmixed.
• Browser source decoding runs through FFmpeg, not the browser’s own audio stack, so your audio device cannot silently choose the import rate. Any output-device resampling that happens during playback is reported separately.

Precision, and what conversion cannot do

• Existing valid PCM16 WAV files are adopted unchanged.
• 32-bit integer and 32/64-bit float WAV sources are supported, and their higher precision is explicitly quantized into the canonical PCM16 listening delivery. FLAC files are decoded at the bit depths the bundled FFmpeg build actually supports; no bit depth is invented where the decoder cannot represent it.
• If your source is itself lossy — an MP3, for example — both chains start from that same decoded lossy source. Audivane never claims to restore information a lossy file already lost, and the desktop and browser codecs are not claimed to produce identical output.

Every prepared side is a canonical PCM16 WAV at the selected session rate, with its own SHA-256 hash and per-stage codec provenance recorded in the evidence view. The source file’s hash always covers the original file you selected.