tooly

Audio

Cut, convert and compress audio and extract sound from video right in your browser. Transcription runs on our servers.

Which tool for which job

  • A device, app or form refuses the format: Convert audio reads MP3, WAV, OGG, M4A, FLAC, AAC, OPUS and WMA and writes MP3, WAV, OGG, M4A or FLAC.
  • The file is too big to send or upload: Compress MP3 re-encodes at 64, 96, 128 or 160 kbps, optionally in mono.
  • Only part of the recording matters: Cut MP3 for a quote, a ringtone or the one usable take.
  • A platform only accepts video: Audio to video puts your track under a still picture as an MP4.
  • The sound is inside a video: MP4 to MP3 drops the picture and keeps the audio.
  • You need the words, not the sound: Transcribe for text or subtitles, Translate video for English text from another language. Both accept audio files as well as video.

Lossy, lossless and what a bitrate means

Every audio format belongs to one of two groups, and the group decides what a conversion can do.

Lossless formats keep every sample. WAV stores raw PCM: CD-quality stereo runs at 1,411 kbps, about 10 MB per minute. FLAC packs the same samples like a ZIP file, typically to 50 to 70% of the WAV size, and converts back to identical audio.

Lossy formats (MP3, AAC in M4A, Vorbis in OGG, Opus, WMA) discard parts of the sound a hearing model judges inaudible. That detail is gone for good. Three practical rules follow:

  • Lossless to lossy, such as WAV to MP3, is the normal direction.
  • Lossy to lossless restores nothing. It only helps when a program insists on WAV, and the file becomes several times larger.
  • Lossy to lossy adds a second generation of loss. Once at 192 kbps is very hard to hear; repeated conversions are not.

For a constant bitrate, the size is simple arithmetic: kbps divided by 8 gives kilobytes per second.

  • 64 kbps is about 0.48 MB per minute, enough for speech in mono.
  • 128 kbps is about 0.96 MB per minute.
  • 192 kbps is about 1.44 MB per minute, a four-minute song is roughly 5.5 MB.
  • 320 kbps is about 2.4 MB per minute, the maximum MP3 allows.

You can work backwards too. A 45-minute podcast of 108 MB is 40 KB per second, which is 320 kbps, far more than a voice needs.

The converter's outputs are fixed for a good balance: MP3 at 192 kbps with the LAME encoder, M4A as AAC at 192 kbps, OGG as Vorbis quality 5 (around 160 kbps), WAV as 16-bit PCM, and FLAC lossless. Sample rate and channel count are kept from the source.

Choosing settings by content

Speech sits in a narrow frequency range and is usually recorded with one microphone. 64 kbps mono suits voice notes, calls and lectures; 96 kbps mono suits a podcast people will hear on headphones. Mono matters more than bitrate here: at 64 kbps stereo the bits are split between two channels and voices turn watery.

Music needs stereo and more headroom. 128 kbps works for background listening and earbuds, 160 kbps for careful listening and for cymbals, strings and reverb tails, which degrade first. Avoid mono for music unless the recording already is mono.

A higher bitrate never improves the source. A 128 kbps MP3 re-encoded at 192 kbps only gets bigger.

What runs where

All four audio tools run FFmpeg compiled to WebAssembly in your browser. The engine is about 30 MB, downloaded on first use and cached. Files are processed on your device and never uploaded, which suits unreleased music, confidential meetings and private voice notes.

Browser limits depend on the tool: 256 MB for the converter and the cutter, which covers several hours of MP3 or about 25 minutes of uncompressed WAV, and 500 MB for the compressor and audio to video. Long WAV files on phones can still exceed the tab's memory.

Convert audio also has a server option for larger files. It runs the same FFmpeg settings, so the result is identical. One server run per rolling 24 hours is free for files up to 500 MB; Premium removes the daily limit and accepts up to 4 GB. The upload is deleted once the result is ready, and the result expires about an hour after the job was created.

Transcription and translation always run on the server, with the Whisper speech model. Free use covers recordings up to 10 minutes, Premium up to 90 minutes per file.

Workflows that chain well

  1. A voice memo for a messenger: cut the silence at both ends, then compress at 64 kbps mono. An hour fits in under 30 MB.
  2. A podcast episode on a video platform: prepare a 1920×1080 cover with Crop image, then use Audio to video. The picture adds very little; an hour at 192 kbps is about 90 MB.
  3. Notes from a recorded lecture: extract the sound with MP4 to MP3 at 128 kbps, trim the parts you need, then transcribe them.
  4. A ringtone: cut 20 to 30 seconds starting on the hook. At 192 kbps, 30 seconds is about 720 KB.

When something goes wrong

  • The output is bigger than the input. Either the target is lossless, or the chosen bitrate is above the source's. Pick a lower bitrate or keep the original.
  • The album art disappeared. Cover images are a separate picture stream, and the converter and compressor keep only the audio.
  • Speech sounds metallic. Use mono at the same bitrate.
  • Music sounds hollow. The recording relied on wide stereo; go back to stereo at 128 kbps.
  • A file is rejected. Some .m4a or .webm files from screen recorders also carry video. Use MP4 to MP3 for those.
  • Audio to video fails. The browser build of FFmpeg could not read the audio. Convert it to MP3 or WAV first.