Speaker 1: Public access to intelligent computing systems like IBM's Watson has opened the door for other tools that make producing journalism a lot faster. For example, Pietro Passerali is both a documentary filmmaker and a software developer. He's integrating the technology into an open-source tool that makes it easy to turn video interviews into edited stories.

Speaker 2: With the tool that I'm working on, you can get a transcription to the IBM Spatial Text Service, and then on that text, you can make selections of the lines that you want to include in your final video, and then export those selections as a video sequence. It provides a user interface that can make it easier for the user to do those selections and explore the text and the video in conjunction one to the other. So if you click on a text and it takes you to the corresponding part of the video, and while the video is playing, the text changes color to show you where you are.

Speaker 1: In other words, it takes about five minutes for an auto-generated transcript to appear, and from that, a journalist can edit a video just by selecting and arranging blocks of text.

Speaker 2: I left it open so that it doesn't matter if you upload an audio or a video file. So in that sense, if you're working as a radio producer or a podcaster or things like that, then you'll still be able to get a transcription.

Speaker 1: Tesla started combining automated transcription and video editing before IBM's Watson was available. But the new IBM tool is faster, and it also has the capability to get more accurate the more it transcribes. Even so, the system still needs a little human help.

Speaker 2: I found that if you're working with video in the use case in which you're just selecting quotes to go straight back to the video interview, it doesn't matter too much if it's like 100% accurate or not, because you just need to get a general sense. It becomes a bit more important if you're working with something like an interview piece or a text article where you want to publish the whole piece. And so in that instance, you take it as a draft, and then you need to go back and fix it. That really varies depending on the accent, the quality of the recording. So if it's a good sound from a good microphone in a production setting, it tends to be more accurate than if it's a phone somewhere else.

Speaker 1: For developers who want to build their own tool with IBM's speech-to-text capability, it's free up to the first 1,000 minutes of transcription. For that, it costs about $1 per hour. There are also limitations on the file format and duration. Pesarelli says speech-to-text is just the beginning.

Speaker 2: There's a whole load of more cognitive capabilities and other things that can be done that is very interesting.

Speaker 1: Similar technology is also available from Google and Microsoft. For the Futures Lab, this is John Doty.