What an OGG file contains
OGG is an open container format from the Xiph.Org Foundation. The container itself does not decide how the sound is compressed. Inside, you will usually find one of two audio codecs:
- Vorbis: the original OGG codec, common in older files, games and some music.
- Opus: a newer codec built for voice and internet audio, used by many messaging apps. Opus files often carry the .opus extension, though the container is still OGG.
Where OGG and Opus files come from
Voice messages are the most common source. Telegram voice messages are Opus audio in an OGG file. WhatsApp voice notes are also Opus in OGG, and they usually appear as .opus or .ogg files when you save or export them. If WhatsApp is your main use, the separate WhatsApp voice note page covers exporting from the app.
OGG also shows up in other places. Firefox records audio-only clips as OGG when a website uses the browser to record. Audacity can export OGG Vorbis, some Linux sound recorder apps save OGG by default, and much of the spoken audio on Wikimedia Commons is stored as OGG.
If the OGG file is already online, for example on your own server or in Dropbox or Google Drive, paste the link into the import box instead of downloading it first. On Telegram Desktop, you can right-click a voice message and choose Save As to get the file.
OGG to text: how the codec affects the result
Opus has a dedicated mode for speech and stays intelligible at bitrates far lower than MP3 needs. That is why messaging apps can send long voice notes as tiny files. Those files are heavily compressed, but the voice usually comes through clearly enough for accurate transcription.
Vorbis was designed with music in mind and needs a higher bitrate for the same clarity. At the bitrates most files use, speech is still well preserved.
In practice, the codec matters less than the recording itself. A voice note recorded with the phone held close to the mouth in a quiet room transcribes very well. One recorded on a busy street, in a car with the windows open, or from across a table will need more corrections. Music, crosstalk and strong accents also reduce accuracy.
Short voice notes have one extra challenge: there is little context around each word. Names of people, places and products are the most likely to come out wrong, so give those a quick check in the editor while the synced player plays the note back.
Transcribing a batch of voice messages
Voice messages tend to come in groups: a chain of notes from a client, a set of interview answers, or a week of reminders to yourself. Each file becomes its own transcript, with timestamps and a synced player. Rename the files before uploading if you want to keep them in order.
The free plan has a daily file allowance. Paid plans allow batch uploads and many more files, which helps when you have dozens of short notes. There is no need to convert anything, because both .ogg and .opus files are accepted directly.
Voice messages are often personal. Files are never used to train AI models, and you can delete any message and its transcript when you no longer need it.
Exports and languages
Speech in 90+ languages is recognised, with automatic detection or a language you choose per file. Messages that mix Chinese and English are supported. Copy the text with one click to paste it into a chat or email, export TXT, Word or PDF for records, or use the Google Translate link if you need to read a message in another language. For audio used in a video, SRT and VTT subtitle exports are available too.