Welcome to Equestria! An online tool to help you with Forced Alignment. This page will explain most of the features Equestria has and how to use these features. Note that you need an account on this website in order to follow the tutorial on this page.
In this module you will learn about forced alignment and experiment with it. Forced alignment is the synchronisation of an audiofile with its corresponding text. Forced alignment will yield the timestamps for every word (and phoneme) in the text that belongs to the sound file. So as input you minimally need an audiofile and a plain text file with the corresponding text. You may (but need not) supply a dictionary file with words and their phonemic transcription for words that are not foreseen in the Forced Aligner itself but this you will find out in this module.
Note that you can go to any stage of Forced Alignment at any time with the buttons in the header below. However, also note that the Forced Alignment process is to be completed in the order presented in the header menu. Later, when you are done with one run of Forced Alignment you will be able to rerun any stage of the project at any time. We will go through the process of Forced Alignment by following the header menu items beginning with selecting a project.
We noticed that you have not logged in yet, please do so on this page.
All the tools that are available on this website require the usage of "Projects". Projects provide a way to store and retrieve files later. Only files uploaded to a certain project can be used in the Forced Alignment tools on Equestria. You can create a project on this page. Please make sure to only use alphanumeric characters for your project name (e.g. [a-z, A-Z, 0-9]) and don't use any spaces in the project name. When you have created a project you will be automatically redirected to the "Upload files" page.
The next page you come across is the "Upload files" page. This pages allows you to upload, download and manage all files within a project. You can upload files by dragging and dropping them in the box on the website. The container is displayed below. If this does not work you can always use the button. The already uploaded files and the files later generated by the Forced Alignment process will show up in the "Files" list below the upload container. Here you can either or all project files.
Another important thing to note about running Forced Alignment is that it consists of several processes chained together. For each process, required files are needed to start this process. A set of these requirements is called a "Profile". Before running a stage of Forced Alignment, you must thus make sure that you have all the required files for at least one Profile uploaded in the Project. The right hand menu of the Upload files page will display a "Profile Configuration" and a "Available profiles" tab. Let's first discuss the "Available profiles" tab.
The available profiles tab will display all Profiles (configurations) for which a certain part of Forced Alignment can be ran. Profiles can have four types of files characterized by two types of options. Files can be:
This thus gives the possibility to four types of files:
In the available profiles tab, a list of profiles is displayed and for each profile all (optional) files are listed. You can thus use this overview to see which files are needed to start a part of Forced Alignment.
For some parts of the Forced Alignment process, you are required to specify the profile configuration before being able to start that part. An example of such a window is displayed below. You are first required to select a Profile in the upper selection container. After selecting a profile the files for that profile will appear. When a file is required, it is displayed like this: Required file*. When a file is optional it is displayed like normal text. When a file is unique it is displayed as a single selector box like the InputWavFile* in the example below. When a file is not unique it is displayed as a multiple selector box like the LexAddOn file in the example below.
When changing profile settings, be sure to wait a little bit before the settings are synchronised to the server. The synchronisation process is indicated with a icon. After the synchronisation is complete you can start Forced Alignment with the button.
Also note that all files have a preferred extension (e.g.
.txt for text files or
.wav for audio files). It is best to select files with the same extension as the profile
indicates as otherwise the Forced Alignment process might not work.
For all parts of the Forced Alignment process, the same interface is used. On the Forced Alignment screen you can see the output of your process under the "Console output" header. The status of the process can be seen next to the "Status" indicator. After the process is completed there will be two buttons on the process page. The button will take you to the next part of Forced Alignment and the button FA process overview will take you to the overview page for all forced alignment processes. Note that G2P is always ran after Forced Alignment, even if your out of vocabulary words. If you want to skip this step you can do so by clicking on the Check Dictionary menu item directly.
The Forced Alignment process overview page will show you all running Forced Alignment processes for the current project. You can go back to those processes by clicking on them. The page also allows you to start a new Forced Alignment process in the same way as was done for the Upload files page.
There are a couple of (non trivial) files that could be outputted by the Forced Alignment process. These files are:
.oovfiles: User lexicon file, can be used to provide add-on functionality for all input textgrid files.
.ctmfiles: A file which has lines with the following information: filename, channel, onset (seconds), offset (seconds), word, confidence score.
.out.tgfiles: A textgrid file with two tiers: word and phoneme. Can be used in combination with original
.wavfile in e.g. Praat for further editing.
After running Forced Alignment the next process that should be ran is Grapheme to Phoneme. This works in the same way as Forced Alignment and this part of the Forced Alignment process also has a process overview on which you can manually start Grapheme to Phoneme processes. The button will take you to the next part of the Forced Alignment process.
Sometimes a Forced Alignment run will not result in success directly. Sometimes the Forced Aligner does
not recognize all words you have put in the transcription file. When this is the case, these words end up
.dict file in the project. On the Check dictionary page you can view the contents of all
.dict files on the server and edit them. Afterwards you can save all files and restart Forced
When your Forced Alignment process is done you will end up on the overview page which will show you whether or not your Forced Alignment process completed. You can go back to any stage of the project or download the project archive with all project files.
This page was last updated on 26-01-2021