SpeakerText Creates Interactive Video Transcripts
Silicon Valley startup SpeakerText today launched a redesigned product that creates interactive transcripts of online videos.
The company transcribes videos using a hybrid of proprietary speech recognition, natural language processing, and crowd sourcing technologies.
Acknowledging that machine transcriptions aren't always accurate, Matt Mireles, CEO of SpeakerText, says the company also employs people to review the transcripts. “There are severe limits to the capabilities of automation, so we added humans to the process,” he says. “Once we get a rough transcript, we run it through the process with humans.”
Once completed, the transcript is placed within a plugin called the SpeakerBar that appears beneath each video that's been transcribed. The SpeakerBar shows users the full video text and automatically highlights each sentence as the video plays. With the SpeakerBar, users can scroll through the full transcript and click on the text to skip backwards and forwards through the video. They can also use the SpeakerBar to search for quotes within the video and share specific moments from inside the videos on their Web sites, blog posts, email, or social networking sites like Facebook and Twitter. When sharing the video transcripts, the SpeakerBar automatically creates a link, called a QuoteLink, that takes people back to the original Web site and starts playing the video at the time cue where the quote appeared.
“Once the page loads, you can take the text and make it interactive,” Mireles says. “You can search it, find the part that your interested in, and jump right to it.”
SpeakerText also analyzes each video transcript to create descriptive metadata. These metatags are placed within the HTML code on the user’s Web site and used to create XML sitemaps so that videos can be indexed by video-specific search engines.
SpeakerText launched a private beta on September 8. Since then, Mireles says the company has been flooded with inquiries from major mainstream and Internet media companies. “The market has been dying for something like this,” says Mireles. “The real sticking point has been the transcription. Unlike purely automated services, we add humans to the process, which lets us guarantee consistent high quality across a variety of video sources.”
The service is compatible with Ooyala and self-hosted videos using the JW player, adding to existing integrations with Brightcove, YouTube, and Blip.tv. In addition, SpeakerText has been redesigned to emphasize sharing video quotes, adding a column with Facebook and Twitter sharing buttons to the left side of the transcript.
According to Mireles, the service is available to anyone, from the smallest blogger to the largest corporation, but ideal customers would include education companies, media companies, and digital archives. Pricing starts at $19.99 a month for the most basic service and varies depending on volume, video length, page views, and other factors.