This project has been moved to https://github.com/smlum/scription
An editor for Amazon Transcribe transcripts!
Amazon Transcribe is a (nearly) free way to automate the transcription of audio files. While the output is fairly accurate, the service can make mistakes. Amazon usefully provide an output json file which contains confidence levels for every word, but does not provide a service to make use of this information.
The AWS Transcription Editor takes in Amazon Transcribe's output files, and visualises them in a browser-based editor. Low confidence words are highlighted to make the transcription process more efficient. Audio files can be loaded and controlled through the interface to help correct mistakes present.
Transcription buddy is an interactive transcription editor. It takes in transcription data files from the automated transcription service Amazon Transcribe, and provides a way to easily check/ correct/ code the output alongside the audio.
The app is probably most useful for people trying to transcribe
These instructions will guide you through using the Amazon Transcribe service and getting a copy of the project up and running on your local machine.
- Have an Amazon Web Servives account
- Have S3 and Transcribe set up on your AWS account
- Install http-server* or use another server
Note: http-server requires node. The recommended way to install node is via the node version manager (nvm):
curl -o- https://raw.githubusercontent.com/creationix/nvm/v0.34.0/install.sh | bash
`nvm install node``
- Create a bucket for your audio in S3
- Upload the audio file
- Set the permissions for the file to 'public access' (read object)
- Copy the url link to the file (found in the 'Overview' tab for the file)
note you may have to set your region to US East (Ohio)) for this stage to work before starting you can also create a 'custom volcabulary' if there are words used in your audio that AWS is likely to misinterpret (e.g. it often hears 'jesus' instead of 'GIS')
- Navigate to the 'Transcription jobs' page
- Create new job
- Fill out form with name, audio url (from the S3 file), adding custom volcabulary, set speaker identification to 'Enabled', and data source to 'Amazon defalut bucket'
- Once the job is finished download the transcription (which shold be json file)
- Clone the repository:
git clone https://github.com/samFredLumley/aws-transcription-editor
- In the "json" folder copy the downloaded json file (suggested name "transcript.json")
- Launch the app on a server. E.g. from the folder containing the project directory run
cd aws-transcription-editor
http-server
You should now be able to access the app from your browser
- Enter the url for your audio and the filename for your json file
- "Load transcript" visualises the json. You can make edits to this text. Changes are autosaved locally but it is recommended to copy and paste out your work regularly.
- You can control the audio while writing using keyboard shortcuts:
Task | Shortcut |
---|---|
go back 5s | ctrl + , |
skip 5s | ctrl + . |
slow down | ctrl + shift + , |
speed up | ctrl + shift + . |