3 ms·

I discovered today that Adobe is introducing Project Shasta, a web-based AI-powered audio recording and editing tool. You can request access now [1]. It seems

I discovered today that Adobe is introducing Project Shasta, a web-based AI-powered audio recording and editing tool. You can request access now [1]. It seems their intention is not only to provide AI-based audio editing and recording but also to cover the ground of what Riverside already offers. Remote recording - Recording with others is as easy as sharing a link. Everyone’s audio is recorded in high quality locally, then Project Shasta syncs it back together in the cloud automatically. "Soon to come features" list things we can already find in Descript, such as filler word removal or speech enhancement. And features already available in the beta include microphone checking for optimal quality and microphone distance, AI-powered audio, remote recording with guests, and project templates. I was waiting to see what Adobe's take on AI-based audio editing and recording was, and it seems it's here. At the moment the project is referred to as Shasta; I imagine these capabilities either format a new product offering or are integrated with existing tools like Audition, Premiere, or After Effects. What do you think? [1] https://pages.adobe.com/shasta/request-invite https://pages.adobe.com/shasta/request-invite

mrandish 4y ago
As an old-school video editor who learned editing on analog VTRs (tape machines) and progressed to digital non-linear editing, I think the time saving potential of AI-assisted audio editing is substantial. I still think it will be a while before AI is ready to create more than a first draft edit but if it handles 95% of the easy cases that's still a huge time-saving - especially on long-form projects. I spent so, so many hours of soul-sucking, rote drudgery doing things like manual audio ducking back in the day.
- dilap 4y ago
  Hopefully something like this catches on w/ smaller podcasters. Listening to shit-quality compressed internet audio in a podcast interviewing someone really interesting is just such a shame!
  - jamiek88 4y ago
    Yeah the volume button dance too. Whisper, whisper, mumble, mumble, EAR PIERCING LAUGH, whisper.
  - anigbrowl 4y ago
    I edit one and clean all that crap out; usually 2/3 of the material ends up being cut, and I've had to write up explainers for journalists on how giving an interview is a different kind of journalism from writing an article. But doing all those micro edits, volume balancing etc. etc. is draining. A 60-90 minute episode can take 10 hours of intense editing by hand, longer if there are quality issues on material recorded in the field. My stuff sounds good but I've been making arrangements to hand off a lot of it to a junior editor and incorporate machine transcription/partial editing to take care of the 'first pass', and I can't tell you how relieved I am.
    - mrandish 4y ago
      You have my sympathy. In doing my own interview recording I've found that starting with good environmental control and properly placed high-quality mics independently recorded at high bandwidth makes a profound difference and saves 70-80% of post labor. I think these AI editing assists could yield another 10% netting an order of magnitude cost reduction for high-quality interview production.

mrandish 4y ago

As an old-school video editor who learned editing on analog VTRs (tape machines) and progressed to digital non-linear editing, I think the time saving potential of AI-assisted audio editing is substantial. I still think it will be a while before AI is ready to create more than a first draft edit but if it handles 95% of the easy cases that's still a huge time-saving - especially on long-form projects. I spent so, so many hours of soul-sucking, rote drudgery doing things like manual audio ducking back in the day.

dilap 4y ago
Hopefully something like this catches on w/ smaller podcasters. Listening to shit-quality compressed internet audio in a podcast interviewing someone really interesting is just such a shame!
- jamiek88 4y ago
  Yeah the volume button dance too. Whisper, whisper, mumble, mumble, EAR PIERCING LAUGH, whisper.
- anigbrowl 4y ago
  I edit one and clean all that crap out; usually 2/3 of the material ends up being cut, and I've had to write up explainers for journalists on how giving an interview is a different kind of journalism from writing an article. But doing all those micro edits, volume balancing etc. etc. is draining. A 60-90 minute episode can take 10 hours of intense editing by hand, longer if there are quality issues on material recorded in the field. My stuff sounds good but I've been making arrangements to hand off a lot of it to a junior editor and incorporate machine transcription/partial editing to take care of the 'first pass', and I can't tell you how relieved I am.
  - mrandish 4y ago
    You have my sympathy. In doing my own interview recording I've found that starting with good environmental control and properly placed high-quality mics independently recorded at high bandwidth makes a profound difference and saves 70-80% of post labor. I think these AI editing assists could yield another 10% netting an order of magnitude cost reduction for high-quality interview production.