Podscripter

Podscripter

Retail shop.

1 follower

Visit website
We've Closed Down Shop Thanks to all our customers for your support!
Podscripter gallery image
Podscripter gallery image
Podscripter gallery image
Launch tags:SaaS
Launch Team
Universal-3 Pro by AssemblyAI
Universal-3 Pro by AssemblyAI
The first promptable speech model for production
Promoted

What do you think? …

Ryan Hoover
Product Hunt
I love how you've automated this by plugging into the podcast's RSS feed. How are you transcribing the podcasts? On a related note, I recently chatted with @goopt about his startup Otter. They've developed impressively accurate voice transcription that's done in real-time using their tech.
Craig Cannon
SupaSleeves
Maker
@goopt @rrhoover Thanks man! We're using Google's speech to text video model. Of the services I've tried, it provides the best results. We then split up the speakers by hand, which ends up being a fair bit of work and is why it's 24hrs instead of minutes 😊. It's a workaround at the moment until diarization (the technical term for it) improves. @goopt if you have any insights on diarization I'd love to learn more!
Davis Baer
Youform
Very cool :) Is it still the same pricing for people with very short podcasts? Like 10-20 minutes?
Craig Cannon
SupaSleeves
Maker
@daviswbaer Thanks! Yeah, keeping it as simple as possible for now to test if people want it. Might switch that up in the future.
Krishna De
@craigcannon That would be great if there were pricing say for each 15 min segment for those under 60 minutes
Craig Cannon
SupaSleeves
Maker
Hey! This project stemmed from SpeechBoard (https://www.producthunt.com/post...) which was a text-based audio editor @ramonrecuero and I made. We found that many people were just using it for automatic transcriptions so I put this together. Let me know what you think!
Calvin Chu
Does effectiveness drop with podcasts with technical jargon/vocabulary/proper nouns, such as blockchain podcasts?
Craig Cannon
SupaSleeves
Maker
@calchulus such a great question. Short answer is yes. How much depends on the speaker's diction. That said, it'd be possible to train a model to just your podcast so it'd know all the vocab if you went back and did some error correction. Proper nouns can be tough, especially if they're mixed with another language. E.g. the sample episode on the site is in english but it's about riding bikes in Italy so many of the proper nouns are off.
Calvin Chu
@craigcannon would the model be something we'd have to provide or a future level of service package? <3 Would love to see how I could help!
Craig Cannon
SupaSleeves
Maker
Owen Williams
Write Together
Been looking at doing this for a long time. I even tried putting my files into Google's Cloud Speech API. It sucked. Instant subscription from me. Looking forward to getting my first episode!
Craig Cannon
SupaSleeves
Maker
@ow haha. Yeah, in many ways this is just a user friendly version of Google's API. Thanks for trying it out!
Cody McLain
How does this compare to Temi.com which is only .10c a minute and also labels the speaker?
Craig Cannon
SupaSleeves
Maker
@codymclain Temi's super cool though in my experience it doesn't get speaker labels right.
Six
ShipJump

I'm just hoping this is a fluke and not a scam.

Pros:

There aren't any. Great idea, extremely poor execution.

Cons:

This says 24 hour turnover...it's been 8 days and zero responses for updates the last two days