A $1.2 million cash injection from New Zealand and Silicon Valley investors will breathe another year of life into local transcription service start-up TranscribeMe, chief executive Alexei Dunayev says.

The Auckland-based company has developed a "hybrid" platform for converting audio and video recordings to text, combining speech recognition algorithm technology - initially developed at the University of Auckland for tracking the sounds of native birds in the Waitakere Ranges - with an army of crowd-sourced human transcribers from all over the world.

Despite being founded only a year ago, more than 100 large companies were already using TranscribeMe's services, Dunayev said.

He said its customers ranged from market research firms to education providers, conference organisers and medical professionals.


New Zealand's Ice Angels investment group, an early investor in the company, joined a number of United States-based syndicates in the latest $1.2 million funding round.

Dunayev said that as well as fresh capital, which will be used to extend TranscribeMe's service to a wider customer base, the new investors would provide the company with crucial access to business networks in the United States.

"By going to the US and raising a professional angel round from Silicon Valley investors we suddenly have access to really smart, really networked, dedicated and driven people," Dunayev said.

TranscribeMe has a team of more than 2500 transcribers, based from New Zealand to North America, the Philippines, Pakistan and Ghana.

Dunayev did not want to reveal how much the transcribers were paid, but said it depended how much time they spent transcribing.

Some earned hundreds of dollars per week, he said.

Many transcribed in their spare time, such as on the bus on their way home from work.

"This really lets people monetise the smallest chunk of their down time," said 30-year-old Dunayev, who was born in the Ukraine but grew up in New Zealand.

The technology works by slicing up raw audio provided by customers into 10 to 15 second "micro tasks", which are put through the computer system and randomly (for purposes of confidentiality) sent out to human transcribers.

TranscribeMe then compares the human and computer-generated transcriptions.

"We try to understand who gives the more accurate answer," Dunayev said.

"The idea is that we're able to do some of the work using computers and of course if we start getting used to your voice the computers will pick it up and get more accurate and people will have to do less and less."

He said only 20 per cent of its transcriptions were at present produced by the technology, with the rest carried out by humans.

But TranscribeMe's technology used artificial intelligence that was continually learning and improving and the day would come when the transcriptions were wholly computer generated, Dunayev added.

He said TranscribeMe's rates - US$1 per minute of transcription for a single speaker and US$3 per minute for multiple speakers - were cost effective when compared with other services.

Customers typically receive their transcription within 48 to 72 hours.

"I think we've created the fastest, most accurate and lowest-cost service that goes from complex voice to word-for-word perfect text," he said.

The company, which has staff in New Zealand and the US, was looking to branch out into new languages, with plans to offer its services to the Spanish speaking world.