I gave up a pretty sweet contracting gig to take on the roll of Entrepreneur. Today I actually feel like one. When you are in consultanting, it’s not the same, even when you have people working for you. Your business scales linearly because the only way to make more money is to find more people or increase your margin (called “the take”) you charge for yourself or for your people. Anyone who knows me can detect that I’m a happiness-over-money sort of person and consulting is only as good as your current client no matter how good the money is. For other’s who are motivated solely by money, well, they can suffer through the worst contracts imaginable but go home with a smile on their face counting their money the whole way. That’s not me though.
This is the first time I’ve removed myself from a contract without any intention of finding a new one. It marks a milestone for me personally, it means I might have to live on Ramen noodle at a time in my life where I should be enjoying steak dinners at a cabin I earned from the proceeds of my career. Instead I’m taking those consulting dollars and betting them on an idea.
Focus is key
I quit my consulting gig for personal reasons, but also because I really suck at doing two things awesomely. I couldn’t in good conscience just “milk my contract” while I try to do the thing I really wanted to do; build the product that’s been rattling around in my head for the past 3 years. It’s like being married to someone but thinking about another person the entire time. My mistress however is this product I want to spend the rest of my life with and I chose to risk it all and go for it. The best part of this risk; and at the same time the most scary part; is that it’s me who determines whether it succeeds or fails, no one else. And really, what risk am I taking here? What’s the worst that can happen? I fail and have to take a 9-5 job? #firstworldproblems if there ever were one.
The idea – Crowdscriber
Croudsourcing is the technique of taking a task, dividing it into pieces and allowing a bunch of people to work on those pieces individually in an effort to finish the task as a whole thing. The idea for Crowdscriber came to me after we (The Basement Coders) did the James Gosling interview. James, being a pretty important person, brought a lot of attention to our little podcast (over 50K hits the first day thanks to Slashdot). The audio quality for this particular cast was pretty bad due to the fact that the interview took place in a San Francisco coffee shop with a lot of background noise. A news site we partner with from time-to-time offered to transcribe the audio for us, in return they’d crosspost it to their site. As it turned out James said some things that went against one of the news site’s sponsors (namely Oracle) and so they handed the transcript over to us for keeps as they didn’t want blood on their hands. It was riddled with errors so I had to painstakingly go through it with a fine tooth comb to fix it all up.
I got first hand experience with how arduous a task transcribing really is. For every 5 minutes of audio I transcribed I’ll bet it took 15 minutes of listening and re-listening. Not to mention the software I utilized for transcription (i.e. an MP3 player and a text editor) didn’t have the UX needed to efficiently transcribe audio. Lots of wasted actions took place like “stop typing, move mouse, scan back through audio, switch back to text editor”.
We also did another interview with some “big guys” at Oracle, and I decided to transcribe that one solely myself. Subsequently it was the last podcast I’ve ever transcribed. Hence the idea for Crowdscriber was born.
What does it do?
I thought to myself “it would be great if our fans could help with transcriptions”. I mean it’s not like every interview we do is in a noisy coffee shop, however the problem with audio/video content is it’s not really noticed by search engines like Google. If the search engines can’t find your content, neither can new listeners.
The basic premise of Crowdscriber is to take an audio/video file, split it up into manageable chunks, distribute the chunks to a group of people (fans, volunteers, employees…), have them transcribe their portion, consolidate all portions and allow fine-tune editing on the whole shebang.
The output will be a transcription of your audio, or a set of subtitles for your video.
I’ve switched focus a bit since the initial concept. The process of accepting audio or video files directly, chopping them up and distributing them amongst transcribers requires storage and a lot of server-side logic for splitting, distribution and management of files. I thought to myself “where would this help people the most?”. YouTube, that’s where! So, to keep things simple while still reaching a large market I changed focus to deal only with online content, specifically YouTube. Much less moving parts.
There are other companies out there that offer transcription, but they are all closed systems. You give them some content to transcribe + $40 for every hour of content and in a few weeks they spit back a transcript. You’ll never know how it was transcribed, or how many errors to expect. YouTube also offers automated transcription, their systems try to do voice recognition. That’s never going to be 100%. It’s actually pretty bad actually. Try clicking the little “CC” button in the lower left hand corner of your video sometime if you need a good laugh.
There is also a “Netflix” competitor called “Viki” in Asia that does crowdsourced subtitles for it’s content. However, it’s not an open system, they use it specifically for their own needs. But it does prove the fact that fans are willing to help you with tasks like transcription.
How Crowdscriber is different
The focus of Crowdscriber is: Quality, leveraging your existing fan-base and multi-language transcription. Humans are basically built for this, computers suck at transcription, the brain is the ultimate fuzzy logic device. It can take multiple impure inputs (re: people’s accents, background noise, etc…) and factor out only the stuff it’s interested in. It finds the signal in the noise.
The business model relies heavily on content providers fan-base, however the option to use a service like CrowdFlower (Mechanical Turk-like service), or our own network of transcribers will be made available as well.
The result will be:
SEO for your Audio/Video properties via transcription & subtitling
Broadening the market to non-english speaking countries via translation
Content publishers on YouTube is our primary market however massive online open courses such as Khan Academy, Stanford, MIT, Coursera will be top of stack as well. The future of education lies in online courseware, and these systems seem to be limited to only a few languages. That’s a shame because their vision is to make education accessible to all walks of life regardless of geography or culture. The people who need them most are probably in countries that don’t speak English. Not to mention, an English speaking person might be interested in content coming from China, India, Japan or other countries around the world, however they’d never know unless a decent transcription/translation was there for them.
Why this idea will work
People who are “fans” of content want to help its creator. It’s true, fans do this because they want to become a part of the content in whatever way they can. They feel a better sense of connection and ownership over the content if they know that they are helping make it a success. In this way we’ll build a loyalty and culture multiplier between content creators and their audience of consumers. We’ll also give the content creators a means to which they can compensate their transcribers by offering them merchandise, gift cards, money or perhaps some sort of “Crowdscriber points” which can be redeemed for all of the above.
So that’s it in a nutshell, there’s more to come! Next stop… Chicago! (you’ll know why in the next posting 😉