Scheduled to be released as a public beta in May, the $249 plug-in hooks into Premiere Pro as a panel and works quite simply. Open the panel, select a timeline to be transcribed and then select a cloud-based transcription service. Transcriptive gives you two options — Speechmatic or Watson. Jim Tierney of Digital Anarchy notes that Speechmatics seems to be more accurate than Watson on poor audio or audio with more background noise.
Prior to any uploads, though, it is necessary to set up accounts with these services. Speechmatics charges .07 per word, while Watson charges .05 per word with the first 1000 words free.
Transcriptive uploads the audio to whichever service you choose and it is then queued for transcription. Processing time is about 2x real time. When the transcription is complete, it will download to the Transcriptive window, identifying different speakers and noting timecode. Accuracy is said to be in the 96% range. But manual correction is simplified since hovering the mouse over a word highlights it for editing.
Once the transcription is complete, it can then be used for metadata, close captioning or even timeline editing. Clicking a word will take the playhead directly to that point in the timeline.
Of course, one might say that sending material to a transcription service will be more accurate. It is certainly an approach when the budget permits. Transcriptive saves time and money at the expense of manually editing copy. Yet the ability to cut video from transcribed text is particularly important for interviews or docs.
Adobe Story performed similar tasks, however, it was designed for scripted scenes. Transcriptive works with any text.
Look for the beta this month.