Do you have spoken-word audio files in need of tagging, transcription and preservation? Trying to start a podcast or share your oral histories online? If so, then Pop Up Archive may be the answer for you. With Pop Up Archive, simply upload your files and then sit back while they automatically apply tags, and transcribe the audio, making it searchable down to the second. It generally takes Pop Up about the length of the recording to create a transcript, so a 30 minute recording should take about a half-hour to process. Pop Up allows users to edit the generated transcripts and their suggested tags.
Use of Pop Up Archive requires signing up for a plan depending on how many hours of audio you will be processing a month. There are personal plans, including a free option, and business plans. The free plan provides one hour of audio processing each month and only basic transcripts for the first two minutes of your audio files, but premium transcripts can be ordered for $22 per hour. Premium transcripts are more accurate and are created using software designed for broadcast and oral history recordings. So far Pop Up can only process English language recordings, but their hope is to support Spanish soon
Another service option is a customizable enterprise plan. This is useful if you have a collection of audio files that need to be uploaded once, rather than the need for recurring uploads. Pop Up Archive receives its support from the John S. and James L. Knight Foundation, the National Endowment for the Humanities, 500 Startups, and Bloomberg Beta.
Users include journalists, media companies, and historians. One major project using Pop Up Archive for support is the Studs Terkel Radio Archive, a joint project of WFMT Chicago, the Chicago History Museum and the Library of Congress. Pop Up Archive provided a customized project plan for the staff of the Studs Terkel Radio Archive and is transcribing the digitized audio files. Terkel was on the radio at WFMT Chicago for almost 50 years and interviewed approximately 5,000 influential and famous people. At this point the project’s website only contains information until the more than 9,000 hours of interviews can be digitized by the Library of Congress and transcribed by Pop Up Archive.
Until the website is complete, materials from the Terkel Archive can be accessed through Pop Up along with other materials from Pop Up Archives public collections. Users can also choose to “Explore” and browse for audio content by tag, creator, interviewer, interviewee, host, guest, and other categories and enjoy all the content made public by Pop Ups users.
I love that this exists, because I love Studs Terkel. Thanks for highlighting it. I listened to Terkel interviewing Tennessee Williams, and started laughing because a transcript was scrolling as I listened, but it was sooooo wrong. Like a bot randomized the interview with other interviews. I would be interested to learn more about the transcription process, and figure out if paying for it would actually be worth it.
I had the same experience listening to his interview with Joseph Campbell. The transcript was so bad it was distracting, but at least for Campbell’s interview, it was just the basic transcription. I am guessing when the project is all said and done, they will use premium transcriptions, which are much better quality.
In their FAQ section, they mention that automated transcriptions are not perfect, but they don’t have to be in order to be discoverable through search engines. So depending on the project, it may not be worth paying for premium transcriptions.
I just listened to a snippet of the Diane Arbus interview and yeah, the transcription was not great. If you’ve ever used a program like Naturally Speaking that translates your speech into text, you know that it takes a bit of training the computer in your voice pattern to get the translation to be reasonably good. And whatever audio translater they’re using hasn’t been trained to specific voices.
But I promise you, when the Pop Up Archive faq says that getting a verbatim transcript that was produced by a human is expensive, slow, and labor-intensive, they aren’t kidding. A rough rule of thumb for the oral histories I’ve been working with is, for a good quality audio recording (which would be a much lower quality than the Studs Terkel tapes, because noise), it takes me eight hours to transcribe an hour of audio. Going along with Catherine’s MPLP comment on my post, it may well be worth it to go the Pop Up Archive route and get a comically bad transcript that can be easily edited by a human rather than get the more accurate human-made transcript **if you’re also getting useful metadata to promote accessibility.** Looking at the Diane Arbus record, you wouldn’t have to listen to a single second of the interview to come up with the tags that they have listed. The “suggested” listed below, I’m not sure how accurate some of them are. But if they’re doing this huge project with the Studs Terkel recordings, presumably they’re getting value from the product that they’re receiving. I’d guess I’d want to see how it worked on material that I was more familiar with. I do like the scrolling transcript in five second chunks though.