(818) 308-4607 Los Angeles, CA
Maintain, optimize and troubleshoot your NLE
Professional cloud workflow platform
Simplified media management
Kollaborate 3.0 is out now on the cloud, which is a major update to our workflow platform that both delivers new features and sets a foundation for future features to build upon.
We're using machine learning technology to create transcripts of the spoken audio in your videos. Kollaborate will automatically highlight the current sentence as the video plays and you can click on sentences to jump to that point in the video.
Kollaborate uses the transcript to create automatic captions below the video and you can even export them as a separate file in common formats like SRT or VTT.
Best of all, this technology is completely self-contained on our servers. Privacy is extremely important to us so we wanted a solution that protected our users' data and did not share it with third-parties. We needed a solution that would also work for our self-hosted customers, some of whom host in environments with no external internet connection.
While it would have been trivial from an engineering perspective to integrate with something like Google or Amazon's speech-to-text technology like our competitors do, those services do not meet this criteria. After a lot of investigation and some code contributions, we finally settled on Mozilla's DeepSpeech. This is a more complex solution to the problem but it gives us maximum flexibility and the ability to finely tailor the technology to fit our customers' specific use-cases.
Our competitors charge extra for transcription and limit the number of hours per month you can transcribe. Using DeepSpeech allows us to make our transcription service free and unlimited. Transcribe as many files as you like for no extra cost and the only limitation is how quickly our servers can process the queue.
Because our speech models are still being refined, we're calling this a beta so that customer expectations are aligned correctly. You can read more details about the situations the model performs best at here, but the short explanation is that content like podcasts and voiceovers will currently return the best results. That's not to say that the transcription feature can't be used in other situations, but the technology is likely to make more mistakes. Over time we will expand the number of situations in which it performs well.
We're building our models on both open source data and data that more reflects our customers' use-cases. You can help improve transcription by correcting any mistakes and then clicking the Learn button. We don't use your data without permission so it is only used when the Learn button is clicked, and you can specify if we can use the entire file or just the sentences you corrected.
The audio of the file then gets cut up into pieces, given a random filename and uses the transcript you provided for training. We don't share your original audio with anyone and once it's part of the model the audio can't be extracted back in its original form.
Another way you can help is by contributing your voice to Mozilla's Common Voice project. This is a public domain speech dataset used by Kollaborate, Firefox and a variety of other projects to create open and privacy-conscious speech technology. All voice data is useful, but especially when it comes from women and non-Americans, who are currently underrepresented in Common Voice's dataset.
We've made many changes to improve the user interface, the most significant of which is the vertical navigation bar which is designed to balance out the interface and place your content closer to the center of the screen.
If you find yourself needing more horizontal space, hover over the Kollaborate logo at the top left and click the arrow that appears to collapse the navigation bar. The navigation bar will automatically collapse by itself on small displays or when you resize your browser window.
While video is the most popular format used on the site, our customers also upload many other file types such as MS Word and PowerPoint documents. The only way to view these files in the past was to download them.
Now Kollaborate has a specific Document file type that supports file extensions like PDF, DOC, DOCX, XLS, XLSX, PPT, PPTX, ODT, ODP. These files will now show thumbnails and be viewable directly in the browser once converted by our servers.
Leaving a comment on a document now tags that comment with the current page number and clicking on a comment will immediately take you to that page.
Columns in List view on the Files page can now be resized or dragged to change their order. You can also right-click to hide them or show additional columns. You can now, for example, show the number of comments a file has next to its name and sort the list by this field.
Advanced Search has been completely overhauled to be more powerful. Use criteria like file size, type or width to narrow down your search. You can even locate files with specific words in their transcript.
Images can now be zoomed and navigated with a lot more control than before. Annotations can be drawn over the image at any zoom level.
Versions can be given custom names like "Rough Cut" or "Fine Cut". To do this, click the purple number next to the filename to view all of the versions, then right-click a version and select Rename Version.
Even though you can leave comments at specific timecode positions and draw over a video, sometimes that isn't enough to get your point across, so you can now attach files to comments. So you can say "I want the color to look like this" and attach a photo, rather than trying to describe it.
Kollaborate is an essential cloud workflow platform that allows you to share files with clients and team members while integrating with Digital Rebellion apps and services. To find out more, see the overview or register for the free trial.