Description

A small Bay Area, angel-backed start-up called SpeakerText provides publishers with a plug-into transcribe videos with clickable and shareable transcripts. The software works with YouTube, Blip.tv, Brightcove and just went live with Ooyala's player and soon with Vimeo. The company creates transcriptions through a process which is part human/part machine, using speech-to-text software, natural language processing and transcriptionists who are part of the Amazon Mechanical Turk labor pool.

Transcript

SpeakerText's Video Transcript Products for Publishers So, what we do is we provide a suite of integrated services for video publishers that makes their content discoverable and social. We do that by turning the video into text. And so sometimes, you would say, “Oh, you guys are a transcription service.” Well, that’s part of what we do but it’s a lot more than that. It’s a fairly simple process from the publisher perspective. But basically, we turn the video into text, host the transcript in the cloud and load it automatically and intelligently under the publisher’s website sort of effortlessly once it gets set up underneath the matching video and then we have a java script application that takes the transcript in terms of interactive transcript, thus you just search and navigate the video. You can tweet out video quotes and those quotes will link back the video at that exact moment inside the video where that quote came from. If a blogger comes and copies a quote from the transcript, that quote automatically links back to your site and starts the video again at that exact moment inside the video. From the publisher’s perspective, it’s really easy to set up. You just connect to your YouTube account, your Blip account, your Brightcove account and then install our plug-in on your website and just tell us which we’ll use to transcribe and then that’s it like you really don’t have to do anything else because it will load automatically server-side into the HTML. And you won't just get sort of a crummy automated transcript which is -- and that’s the problem with sort of YouTube automated captions in a ramp. Ramp solution is that the purely algorithmic, the purely automated solutions, they only work if you have perfect sound quality in a single speaker speaking in the sound studio with a perfect voice. That’s a small percent of the market and that’s great for that. But the problem with that is that anytime you get any sort of variants from that kind of norm, the results kind of fall off the cliff. And so, what we do is we combine artificial and human intelligence. We sort of start off to speech recognition and then we have a human layer on top of that that starts from maybe this rough transcript and turns it into something really accurate. We started out building off of Amazon Mechanical Turk which is a crowd source labor pool that people can tap into. What we've done is we’ve built sort of an intelligent quality assurance process on top of that. Part of our technology is machine learning algorithms that track worker history, identify good workers and distinguish them from bad workers and weight them differently. So, it’s a multi-step process. So, think of it as an assembly line and like a modern assembly line, some parts are automated, some parts have humans. The beauty of this system is that from an artificial intelligence perspective, machine learning perspective, we are able to sort of capture all these data, all these transcripts and feed them back into our algorithms and make them smarter and more intelligent over time. It’s a premium service and what we want to do is we want to be able to service kind of the whole market. So, you can actually sign up today on SpeakerText.com and we actually have transparent pricing where there are different pricing tiers. And so, it is a hosted service. It’s not just pay by the minute transcription which I think something people get confuse on a little bit. And so, the lowest package starts off at $20.00 a month and then you just pay per minute in transcription and it goes up from there depending on how many video page loads you’re doing. And so, the self-serve packages range from $20.00 a month up to $1000.00 a month and then beyond that, if you go above a half a million video page views, that’s an enterprise deal and we negotiate that directly. So, if you have PayPal account or credit card, just sign up out of the box. It works with Brightcove, YouTube, Blip.TV, very soon Ooyala and then after that, Vimeo. If you're using your own player or one of this open source players, we’re actively working to integrate with them. Once you install the plug-in on your site, we have a service-side script that loads the transcript into the HTML but part of it is JavaScript. And so, it loads client side. What it does is you just embed the video on your site as you normally would and then speaker text loads the transcript directly underneath the video and it’s in this collapsible interactive transcript player. So, if you want, you can just have it load in sort of a collapsed little sort of bar that will say, expand transcript. When a person clicks on that, they can open the transcript drawer, read the whole thing, click around on the text, search within the video; share video quotes on Twitter just by highlighting the text. If they copy text into a blog post that quote will then automatically link back to the publisher’s page and start the video at that exact moment inside the video. So, it’s great for consumers and for the end-user but it’s great for publishers too. SpeakerText is easy to install and works with popular publishing platforms like WordPress and video sharing sites like YouTube, Blip.TV and BrightCove. To learn more, visit SpeakerText.com.