Awesome to see this make it to public release! I've been using it for a few days now and it's the best agentic experience I've seen so far so congrats to the entire team!
We've all been on both sides of this article. Sometimes (especially in startups) we overcommit and fail to hit our deadlines. This can burn relationships and do such harm that I think it's an incredibly important skill to develop, particularly for entrepreneurs.
This skill seems to go hand in hand with the ability to say "no" when you have too much on your plate.
As a side note... the Edittress platform is pretty cool! It looks like a much needed mentorship platform that I think will receive a warm welcome into the tech community particularly now that we are having more open discussions about mental health, inclusion, and diversity.
I hope we can expect a ShowHN soon! I just signed up.
2. The app itself is on Rails with AWS Polly transcribing most of the requests. x% of the transcriptions are routed to a Flask API that runs a neural TTS engine + with token-based SSML rolling out slowly for some heavy-users.
3. I have no plans to monetize Polly Podcast.
4. Unlimited readings.
5. I'm working now with large publishers to figure out what the costs look like to generate new synthetic voices at scale. Hope to release this publicly as soon as I get the go-ahead.
If you have any questions feel free to reach out in email =]
I’ve been working on Polly Podcast for a few months and I’m excited to share it with the community. I made this so that I could listen to articles while at work or during my commute.
Once I got into building the platform I found this incredible voice synthesis community in NYC, which led me to Recess Labs[0], which is where I am now while I work on building my own voice model based on the recent Tacotron 2 paper[1].
Would love feedback and I’m happy to answer any questions or talk about voice-tech.
I've never needed to subtitle a video so I don't really know who your competitors might be but none of the links posted so far in this thread (except for yours) seem to be very easy or intuitive to use.
For OpenCV classification tutorials this is another great resource to keep playing around with DIY projects. FYI avoid his email list unless you enjoy 3-4 sales emails every week.
I recently bought a dedicated Moleskine to record algorithms and design patterns I use at work and university. It's been really fun to record multiple solutions and then be able to review my writing several days or weeks later to note an optimal solution.
Looking for formatting tips similar to Bullet[0] journaling, but for more STEM-related notekeeping.
I've had an absolute blast working on bots this year thanks to all of you on the Dialogflow(API.AI) team. Thank you for all the hard work and building such an easy to use product.
Yes, this is true but it depends how you approach the problem.
There are a lot of medium to large retail companies looking to front the cash for startups willing to "move fast and break things". The catch is that they will disown you when you break something.
I'm interested in making this into a framework and it's been difficult to say the least.
The high level things about a startup are very consistent and good candidates to be fit into a framework. I'm not sure but I think these are just called unit economics. Things like "Our product is sold for $X because it saves our customer Y in time". However it's the unknown, and emotional values that I've found incredibly hard to fit into a framework. I'm fascinated now in seeing if the way the finance industry calculates risk is in any way a good framework for startups to assess which features to build.
What I've found to be the hardest questions to answer...
How do I find industry metrics on an industry that doesn't exist yet?
How do I prove people will want something that they don't want right now?
How do I measure emotional value? Tactile sensation(hardware)? UX, UI, etc...
These are the things that startups often believe to be their advantage over the competition. "Our product is much more fun/easy/fast to use/learn/teach" But how do we measure that?
The most valuable exercise I've come up with is this question...
How would your user recreate your product, if you're product didn't exist and they had to piece together the end result with existing technology?
I have a startup right now that creates custom educational podcasts by summarizing publicly available content (with attribution) to generate entirely new content and sort the corpus in increasing complexity. If you searched "Skateboarding", you would get a text document that taught you what skateboarding is, then the history of skateboarding, and then get into beginner, intermediate, and advanced skateboarding lessons. This would go through text-to-voice and be downloaded to your device for offline listening.
Search any topic and you get a 45 minute podcast to listen to on your commute.
In our case, I stepped back and said "Okay... If I wanted an educational podcast on skateboarding, the first thing I would do is search Google, then Wikipedia, then I would start going to skateboarding blogs and read them one by one in increasing complexity. If I wanted to consume this content during my commute I would take all of this content and copy it into a text-to-voice service, and download that audio file on my phone for listening offline." I walked through this entire process and it took me 1 hour to get 45 minutes worth of audio content.
Peter Thiel says that your solution must be 10X better than the existing solution.[0]
Unfortunately for me, I think I will need to cut the time it takes to manually create a podcast by 1/10 and also 10x the quality of the content, which I don't know that I can do.
Tangentially, I often joke that if the problem your startup is trying to solve doesn't exist as a meme, than it's not a real problem for enough people.
I don't think the Tinder model and/or Tinder UI is a good fit for your case.
I'm hesitant to "like" a restaurant suggestion because if I was to match, then I'm stuck going to that restaurant for dinner when there is possibly a better match that just had yet to be suggested to me. The original Tinder model works because there is no risk of being locked-in (lol).
That being said, I think this is a great idea and IMO your initial use-case and test users should be startups trying to decide what to order the team on Seamless. Print out some flyers that say "Team can't decide what to order for lunch?" and post them around co-working spaces.
If I was building this app my first iteration would have been to show the users a list of the top 10 nearby restaurants weighted by stars on yelp and filtered by price/distance. Let them give a 1-5 thumbs-up rating of as many restaurants on that list as they want and total up the number of upvotes for each option.