The average video on a website has a half-life of six days. This means that 50% of the videos lifetime views will happen in that time frame. After twenty one days over 75% of views will occur. As content creators, these statistics are daunting. The content of a video can in many instances provide amazing sound bytes and SEO friendly lines that unfortunately never get tracked. Videos are reliant on titles and tags to convey message and pull. So what, as producers, can you do to make your video content searchable on the web? Take a look at Speakertext.com.
Speakertext.com, a service founded by Matt Mireles, Matt Swanson and Tyler Kieft, turns your videos into text and allows that content to be searchable by Google. I had the opportunity to speak with Matt Mireles, CEO, about the service and why Speakertext.com will become the webs default transcription service.
Transcription technology has been around for many years. The biggest problem with this technology, it is not perfected yet. If any of you use Google Voice and try to read your transcribed voice mail messages than you know the issue. I said “do we need milk?” and it reads “want me to tie you up in silk”. Very different messages. The unique structure of Speakertext.com combines that AI element with human crowd-sourcing. I compared it briefly to what CardMunch does for business cards. The videos are sent through the transcription service then chopped up and sent to actual humans to be refined for accuracy. These sections are then reassembled and a high quality transcription added to your video. Matt and his team are focusing heavily on quality and the scalability of the service. An interesting point to note about crowd sourcing is the difficulty in dealing with humans. By nature humans are smart and crafty. There will always be a level of scamming that Matt and his team will have to deal with as SpeakerText grows. “Transcription technology is still way off from providing the level of quality crowd-sourcing can add.”
Other than the technical nuances of this service the most impressive part is the social layer built in to the SpeakerText bar. If you are watching a video, that has SpeakerText built-in, you have the ability to select text for social sharing. “You can tweet or post to Facebook excerpts from the video in text form”. If someone clicks the link you shared, the video opens right to that point where that text is being said. As well, you can fast forward or rewind the video by transcribed section and search within the transcription for keywords or phrases and play right from there.
Speakertext.com has raised some funding and will disclose in the coming weeks how much and from which investors. Keep an eye on Matt and his team. Their corporate culture is what Matt describes as “scrappy, self made, up and comer types”. SpeakerTexts combination of AI and human interaction will set them apart in the video transcription category and their hope is to become the Twillio of the space.
Do you have a video on your site? How sweet would it be for your video content to be searched?