All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Changed option
--max_words_per_line
to--max-words-per-line
- Fixed utf-8 decoding of dictionaries on macOS
- Improved alignment algorithm in
linennan
. ali
file format compatibility withadskrivan
andlinennan
.
- Fixed model selection when outputing to a txt file.
- Fixed inference with progress bar for MKV files.
- Fixed MacOS compatibility issue when downloading models.
- Better segmentation when decoding to a text file with
-o/--output
option. - Shows a progress bar when decoding to a text file with
-o/--output
option. - Included Anaouder version in EAF headers.
- Removed optional use of translation dictionaries.
- Updated Vosk STT model to v0.9.
- Auto-downloading of models online if not available locally.
- Progress bars with
tqdm
.
- Automatically converts to wave and keep fillers when creating 'split' files.
- Fixed argument propagation error in istitlan.
- Fixed missing module
pytz
- Export to
EAF
(Elan) files. - Option
--autosplit
(adaptative splitting of audio file at silences). - Option
--set-ffmpeg-path
to define the path to ffmpeg explicitely. - Option
-v/--version
to show the version of the module.
- Fixed dependency in
normalizan
script.
normalizan
command will normalize a text file or whatever is sent through stdin.- Updated STT model to v0.8.
- Added option
--keep-fillers
that allows verbal fillers ('euh', 'ebeñ', 'kwa', 'alors'...) in transcriptions.
- Changed option
-d/--translate
to--translate
because of conflicting '-d' option inmikro.py
. script. - Bufixes in script
mikro.py
from @Tornaod
- Fixed setup.py to include non python file, broken by the new folder hierarchy.
- Options
-o/--output
(output file) and-d/--translate
(translate words using a tsv file) added toadskrivan
,istitlan
andmikro
scripts. - Simplified folder hierarchy.
- The scripts can now be called with the python interpreter.
- First version uploaded to pypi.org