Digital Anarchy today released Transcriptive, its new automated transcription plug-in for Adobe Premiere Pro. Leveraging technology from AI and speech-recognition specialists including Speechmatics and IBM Watson, Transcriptive wrangles the process of sending audio to the cloud for transcription purposes inside a Premiere Pro panel so editors don’t have to leave the NLE.
A Second Wind for Speech Recognition
Automated transcription has long been a kind of Holy Grail for video editors, who have been frustrated at the state of the technology to date. Avid first introduced its ScriptSync tool based on technology from Nexidia back in 2007, but pulled the feature from Media Composer in 2014. At roughly the same time, Adobe removed speech analysis features from Premiere Pro.
More recently, Avid reintroduced ScriptSync and PhraseFind as paid options for Media Composer. At the same time, new competitors have come online offering speech-recognition as a cloud-based by-the-minute service for developers. That’s when Digital Anarchy saw the opportunity to develop a tool that would help video editors interface with a new breed of transcription tools.
“The technology has really matured over the last year or two,” Digital Anarchy founder and President Jim Tierney told StudioDaily. “Even YouTube’s auto-transcription has gotten much better over the last year. That’s partly where Transcriptive came from — us wanting to add transcriptions to YouTube videos and going, ‘God, this is awful. There’s got to be a better way.'”
Watch, Edit, Export
When a file is transcribed, users see the text in the Transcriptive panel, which can be floated in front of the workspace or docked among other Premiere Pro panels. A host of keyboard shortcuts are available for editing and punctuating the transcript, and a column on the left-hand side identifies different speakers. Once you get your transcription into shape, verifying its accuracy and making necessary edits, you can choose from a number of export options, including plain text, JSON, SRT and STL, as well as the ability to export to Adobe Premiere Pro clip markers or sequence markers.
We fired up the trial version of Transcriptive, which is limited to two minutes of speech analysis, and let both Watson and Speechmatics analyze the opening of the closest project we had available, our recent podcast with the editors of Detroit. It’s not exactly a torture test, but the results were pretty impressive, requiring only some minor edits for accuracy’s sake. Even celebrity call-outs including “Kathryn Bigelow,” “Aaron Sorkin” and “Ben Affleck” were recognized and spelled correctly.
Tierney says versions of Transcriptive for Final Cut Pro and Da Vinci Resolve are likely on the way. Premiere made sense as a launch platform because it already has a panel API built in. “That allowed us to built in a full HTML5 web app running within Premiere,” he explained. “Now we can take advantage of all Premiere’s other tools. All the ability to handle metadata is in there, so we can drop the transcript into that panel and let you search that. As you’re going through and editing, you can watch the video and the audio as you’re looking at the transcript to make changes. And as soon as you click on a word, you’ll jump there on the timeline. It’s a really effective way to search video for words and phrases.”
Also coming up are different localizations. For example, one of the different languages supported is Japanese, so Tierney hopes to get a Japan-friendly version up and running in the future.
“We’re excited by a lot of cool stuff that AI is making possible,” Tierney said. “As we look toward the future and see these new products coming out, we’re looking at how we can bring the tools into video applications to make video editors’ lives easier.”
Transcriptive users must sign up for Speechmatics and/or Watson, which charge by the transcription minute, but the fees are modest, and free transcribing minutes are available — Watson, for instance, allows up to 1,000 free transcription minutes (sans speaker identification) per month, Digital Anarchy said. Transcriptive itself is sold for a one-time charge of $299.
Transcriptive will be on demo in the Speechmatics booth at IBC 2017: Hall 8, stand C23.