Using Quirkos Transcribe
To add a transcription, you need to be logged in, and have a project open (create a new one if you haven’t already). Then click the main menu in the top right, choose Add Sources, then Transcribe. The video below demonstrates the simple process:
You’ll be asked to confirm your account password (not project password if one is set), as this is used to encrypt the audio file, and collect the transcribed text. You’ll then be shown a file selection dialogue for your platform, allowing you to choose any supported format for your audio or video file. This includes most common formats, such as mp3, wav, ogg, avi, m4v, mp4, and most different types of audio encoding. If video files are selected, the audio file will be removed and recoded, and only the audio uploaded for transcription.
Once the file is selected, you will get a dialogue to set some parameters. By default, the system will guess the language used (by looking at the first 30 seconds of the audio). However, if this does not work, or you want to specify a language, the 5 best supported languages can be manually chosen. Note that at the moment, while more languages will technically work, error rates are much higher than for the 5 languages you can specify – which are English, Italian, German, Spanish and Portuguese. These are the languages which have the best accuracy ‘word-error-rates’ below at least 5% (although usually much lower). There’s also an option to select ‘High-accuracy’ transcription mode. This is off by default, as the standard setting is usually very good. However, if you’ve tried a transcription and got many errors, you can try again with the High accuracy model. This will at least double the time taken to process the transcription, but does provide better accuracy with technical language, noisy environments, and with different languages. For most transcriptions the default options will be fine.
Clicking on the Transcribe button then starts the re-encoding, encryption and uploading process. Entirely on your own computer, the software will open the file to be transcribe, convert it into the correct audio format for the service, and encrypt it with the account password. This process will take several minutes, depending on the speed of your computer, and the length of the file. This means that the audio file is securely encrypted before it leaves your computer, ensuring it is very difficult to open if intercepted without the password. The time taken to upload the file will depend on the speed of your internet connection, but as only a low bit-rate audio file is sent (even if the original file is video) the process usually takes only a few minutes.
When this stage is complete, you will get a success notification, and a rough estimate of how long the file will take to be transcribed, based on the length of the file, and how busy the transcription server is at the time. It operates a ‘queue’ system, so if other people have files waiting to be transcribed, those will be transcribed first. However, we do not anticipate the system to have long waits in these scenarios. While you can only send one audio file at a time to be transcribed, you can add as many as you like in the queue, which will be transcribed in order.
If file uploads fail at this stage, it is likely because the audio format of the file is not supported, and re-saving the in a different format will likely help (especially mp3 or ogg). There is also a 2GB file size limit. You can close your browser and logout as soon as the upload is completed, and the transcribed file will be ready for collection whenever you log back in again. However, you can also keep working on your project, or change projects, but note that results are only imported into the project the transcribe was submitted to.
You can check the status of the transcription once submitted in the ‘Add Sources > Transcribe’ menu, which will show a list of all pending, and completed transcripts awaiting collection for this project. This will show either Pending, In Process, or Success. You will need to manually check this menu to see when the completed transcript has a ‘Success’ status, and can be collected by clicking on that file name. You will then be asked to provide your account password again, so that you can de-encrypt the transcribed text, and add it as a source by your project. This ensures that the text is stored in a fully encrypted state while it is waiting to be collected. For confidentially and security reasons, uncollected transcripts will be deleted after a week, but you can resubmit the file at anytime.
The transcribed text will then be added to the project as a source, which you can review and edit within Quirkos and start coding on right away. You can also export the text, either by copy-pasting, or choosing the Export as Word or Excel options. It’s yours, and there are no restrictions on use!
For the transcription service to work, you need to have ‘minutes’ available on your account. All users of Quirkos Cloud will get 100 minutes free to try the service, beyond this users will need to add the transcription service using their account page when logged in, where the remaining number of minutes that month are displayed. 3000 minutes are credited each month, which allows for 3000 minutes of audio to be transcribed. This does not carry over, and the period is dependent on the subscription renewal date. When you have used your minutes for that period, no more audio files can be uploaded – note that if you only have 40 minutes left, you cannot upload a 60 minute audio file – the whole file will be rejected. You can cancel the transcription service anytime when you no longer need it, and start again at any time. The minimum period is 3 months, and there is a discounted rate if subscribing for a whole year.
For the moment the transcription service only works in the Quirkos Web browser based tool, however you can use this just to do the transcription, and go back to the Desktop version if you prefer as soon as the transcription is submitted and collected. At the moment, Quirkos Transcribe does not automatically identify multiple speakers, or allow you to add recurring technical terms that might not be recognised. However, we are hoping to add these features in the future.
If you have any problems, feedback or questions about the transcription service, please contact us at support@quirkos.com