Wow, Skye! What an incredible article. I was spellbound. I would just like to share my own quick experience with Otter.ai transcription which I was in love with to transcribe my podcasts and videos until they doubled their fees during the pandemic which forced me to find another service and Whisper came along and was completely free which was one way for them to access more data for scraping. You have brilliantly pieced all the players together. I'm looking forward to more on this.
It's GPT, not GBT, but regardless - as a new Descript paid user this is highly concerning. I understand the need for more training data, but there has to be a better model for obtaining it moving forward. People don't want their work to be stolen, and with good reason, and none of this constitutes fair use under any definition of the term. Thanks for reporting on this!
I use Descript and work in information security. I didn't even consider how Descript was doing transcription. Thanks for bringing this to our attention.
Speaking of the Open AI Start-up fund, check out what happened to its Converge-2 program which was announced with great fanfare in Dec 2023, took applications from start-ups, and then.....radio silence. Nobody who submitted heard anything. Then, we found out about Sam's involvement. https://openai.fund/news/converge-2
I'm struck by the dichotomy between Milky Way Galaxy-class technical talent operating in an organization that's as transparent and drama-free as a middle school lunchroom.
Interesting; I wondered about that announcement on the website but made the assumption it had gone as planned. If you want to talk further, send me an email at skyepillsatwork@gmail.com.
Did you know that Hindenburg’s transcription is fully contained and runs 100% on your computer? Your content is never sent to a server and the transcription engine is fully self-contained so there is no possibility of using it for training.
Hindenburg’s transcription does not provide true word-token–level timestamps like you’d get from models such as OpenAI Whisper, WhisperX, or Gentle. So no way to cut words or edit text like Descript. TimeBolt runs waveform detection algorithms local for a cut 20x the accuracy of transcripts. To cut um's, repeats, and bad takes, you need token level alignment, and why waveform cuts are combined with AI transcripts from Amazon Transcribe.
Great article, Skye! First comes the free tier to capture creators. Then the $50M round for AI expansion. And now Descript just boosted pricing. Multi-cam projects that used to count as one timeline now count by number of cameras.
You can build local video automation software, edits faster than real-time, and only transcribes when you ask it to. But TimeBolt.io model doesn’t fit a $2T AI fantasy.
Wow, Skye! What an incredible article. I was spellbound. I would just like to share my own quick experience with Otter.ai transcription which I was in love with to transcribe my podcasts and videos until they doubled their fees during the pandemic which forced me to find another service and Whisper came along and was completely free which was one way for them to access more data for scraping. You have brilliantly pieced all the players together. I'm looking forward to more on this.
Ahhh that's fascinating — yep, that's the trade we don't even realize we're making! Thank you for sharing this story!
It's GPT, not GBT, but regardless - as a new Descript paid user this is highly concerning. I understand the need for more training data, but there has to be a better model for obtaining it moving forward. People don't want their work to be stolen, and with good reason, and none of this constitutes fair use under any definition of the term. Thanks for reporting on this!
Your welcome — and thank you for flagging my typo; it's been updated!
At least with this typo, it's clear that I'm not a writer who ever uses ChatGPT to do my work for me LOL
I use Descript and work in information security. I didn't even consider how Descript was doing transcription. Thanks for bringing this to our attention.
Of course, John!
Speaking of the Open AI Start-up fund, check out what happened to its Converge-2 program which was announced with great fanfare in Dec 2023, took applications from start-ups, and then.....radio silence. Nobody who submitted heard anything. Then, we found out about Sam's involvement. https://openai.fund/news/converge-2
I'm struck by the dichotomy between Milky Way Galaxy-class technical talent operating in an organization that's as transparent and drama-free as a middle school lunchroom.
Interesting; I wondered about that announcement on the website but made the assumption it had gone as planned. If you want to talk further, send me an email at skyepillsatwork@gmail.com.
Did you know that Hindenburg’s transcription is fully contained and runs 100% on your computer? Your content is never sent to a server and the transcription engine is fully self-contained so there is no possibility of using it for training.
Hindenburg’s transcription does not provide true word-token–level timestamps like you’d get from models such as OpenAI Whisper, WhisperX, or Gentle. So no way to cut words or edit text like Descript. TimeBolt runs waveform detection algorithms local for a cut 20x the accuracy of transcripts. To cut um's, repeats, and bad takes, you need token level alignment, and why waveform cuts are combined with AI transcripts from Amazon Transcribe.
Great article, Skye! First comes the free tier to capture creators. Then the $50M round for AI expansion. And now Descript just boosted pricing. Multi-cam projects that used to count as one timeline now count by number of cameras.
You can build local video automation software, edits faster than real-time, and only transcribes when you ask it to. But TimeBolt.io model doesn’t fit a $2T AI fantasy.