Due to bot attacks, we have added this captcha function. To bypass the captcha and gain access to advanced features, we recommend you create a free account. Note that sevens and ones often look the same.
yourzip.zip
├ yourtrans.xlsx/tsv/txt
├ file0001.wav
├ file0002.wav
├ file0003.wav
├ …
└ file9999.wav
Transcriptions in a master file absent of time stamps - as separate rows with separate audio* files for each transcription.
yourzip.zip
├ yourtrans.xlsx/tsv/txt
├ file01.wav
├ file02.wav
├ file03.wav
├ …
└ file99.wav
Transcriptions in a master file with start and end time stamps with more than one row per audio* file.
yourzip.zip
├ file0001.lab
├ file0001.wav
├ file0002.lab
├ file0002.wav
├ file0003.lab
├ file0003.wav
├ …
├ file9999.lab
└ file9999.wav
Transcriptions as separate same-name lab and audio* files, absent of time stamps.
yourzip.zip
├ file01.TextGrid
├ file01.wav
├ file02.eaf
├ file02.wav
├ file03.tsv
├ file03.wav
├ file04.xlsx
├ file04.wav
├ …
├ file99.txt
└ file99.wav
Longer transcription files in TextGrid, eaf, tsv, txt, or xlsx format with same-name audio* files.
*Audio codecs accepted: WAV, MP3, AVI, M4A, AC-3, AIFF, ALAC, FLAC, M4R, OGG, OPUS, WMA
A single upload may be no larger than
25 MB. If you wish to align batches as large as 150 MB,
skip the tedious captcha, and
take advantage of features like the customized pronunciation
dictionary,
create a free account.
If you need help, select
change transcription mode for video guides.
Autophon is a web app that offers free online forced alignment, transforming an audio file and its corresponding transcript into a time-aligned phonetic annotation readable in Praat. Forced alignment refers to the process by which a computational model identifies the time intervals in the audio that contain each phonetic segment. Currently, Autophon utilizes The Montreal Forced Aligner 1.0 for its backend with language-specific models trained on both read-aloud and spontaneous speech. Users can access detailed guides by selecting a language from the list on the right. How does it work?
Don't have an account? Sign up here
Name | Language | Tiers | Size MB | Words | Missing Words | Last status | |||
---|---|---|---|---|---|---|---|---|---|
No data available in table |
25-02-17 Autophon will shut down indefinitely on April 30 due to lack of funding.
Users may contact Nate Young for a free GitHub clone to run locally.
It was a pleasure serving the phonetics community!
Best wishes,
Nate Young and Kaosi Anikwe
25-02-17 Autophon will shut down indefinitely on April 30 due to lack of funding.
Users may contact Nate Young for a free GitHub clone to run locally.
It was a pleasure serving the phonetics community!
Best wishes,
Nate Young and Kaosi Anikwe
24-10-31 We have discovered difficulties with orthography that uses Unicode combining marks instead of their precomposed counterparts. Using, e.g., a with U+0301 for á is now fine because we have added some logic that transforms it to U+00E1. However, using a with U+030B for a̋ will result in the accent being stripped from your orthography because there is no precomposed equivalent in Unicode. We are working on a solution but anticipate late 2025 at the earliest for its roll-out.