diminikolaou

Yeah Postgres is a great example, it comes up often in HN.

We have a small map of tweaks, and our users keep feeding us with more. The model performs great on its own most of the time though.

Making a GPT + text to speech wrapper is not complicated.

Providing all other features (e.g video generation, podcast publishing, auto translation and many other features we’ve added that allow for higher quality pod creation) increase the level of complexity for reproducibility.

Ultimately, we aim to keep building features that lead to higher quality pods, easier to build, and integration of ansiliary (video, translation, show notes) that will enhance our moat.

The user holds all rights. Feel free to read

Terms of Service: https://wondercraft-ai.notion.site/Terms-of-Service-fbfc043c...

Privacy Policy: https://wondercraft-ai.notion.site/Privacy-Policy-f35636350f...

Yeah it's a valid point, podcast consumption is definitely synonymous to commute.

I'd say these are some reasons for continued increase: - Commute has not fallen as much as we might think - More quality content is becoming available, bringing in more audience - Natural expansion from people enjoying the experience and sharing +++

We're lowering the barrier to create a podcast so that might happen to an extent, but on the flip side, valued written content that has been confined in text format can now find its way to audio format. Paul Graham essays is a good example: https://app.wondercraft.ai/p/myqXnyUD

Thanks for sharing those. We're currently testing out the release of personalised podcasts, as the combination of LLMs and the realistic text-to-speech can lead to a solution of a quality level that has not been available until now.

Descript is an excellent "podcast editing" tool, but is not a "podcast creation" / text-to-speech first platform.

It offers "Overdub", but they suggest limited use of it (replacing 1-2 words), not full podcast creation. The quality of the text-to-speech model they use is not at the same level as the ElevenLabs model we are using, so I think someone would struggle to create an engaging podcast using Overdub. I'd also think that this is not a direction that Descript would be heading to, as they have built a successful business as a "podcast editing" platform and catering to "podcast creation" would be a significant change.

In addition to this, given that we are a "podcast creation" platform rather a "podcast editing" platform, we have focused on integrating functionality that allows users to go end to end in creating a podcasts. Examples: publishing a podcast (RSS), generating a video automatically, providing a shareable podcast page, autotranslating the podcast to other languages and a lot more.

Thank you for flagging and sorry about this. Fixing asap, but please use this to access for now https://www.wondercraft.ai/pricing

Yes, but it will take a few months. Looking like Q4.

Thanks @colesantiago - great to hear. That's what we're trying to solve, the effort behind creating a podcast is just too much right now and a lot of content remains confined in text format.

Yeah you can have multiple voices on a single podcast in the current version.

You just need to add more than one section, and then alternate the voice you choose to talk.

Annotations for emotion are coming by the end of August.

Yes - this is full list. English, Spanish, Portuguese, German, Italian, French, Polish, Hindi, Korean,, Dutch, Turkish, Swedish, Russian, Indonesian, Vietnamese, Filipino, Ukrainian, Greek, Czech, Finish, Romanian, Danish, Bulgarian, Malay, Hungarian, Norwegian, Slovak, Croatian, Tamil

Compute to get this quality of text to speech is still quite expensive. Prices will drop as models become smaller. If you need more for now, we do offer Enterprise plans.

Thanks for listening @vishnuharidas! Any feedback always welcome and if you want more languages let us know which ones.

thanks Magdy, there's 30 languages available on the app for anyone to create their podcast in. If there's enough interest for HN Recap in other languages we'll launch those!

Which language would you want it in?

Not on youtube, but check out Hacker News Recap podcast. Daily podcast of the top posts on Hackernews.

https://www.wondercraft.ai/podcasts/hacker-news-recap

For what is worth, most of these words are still being used in modern Greece (I'm Greek).

Every day of the past month, a script has been generating a podcast that recaps the top Hacker News posts of the previous day. The content is modified using LLMs and the audio synthesized using TTS. The response from the HN community has been amazing, so we wanted to share some insights.

Check out more at wondercraft.ai Listen to this post: https://www.buzzsprout.com/2170103/12788630

As long as the community is interested, we'll keep running this. We've seen some strong interest until now, with more than 1000 downloads for the first episode and it grows day by day. Subscribe in your platform of choice so we can keep this going!

HN Recap is a daily show that summarizes the top HN posts from the previous day. In response to community feedback, this episode now also features a summary of the community’s reactions to each post.

The show is produced by Wondercraft, a YC startup, that utilizes AI tools to convert text into studio-quality podcasts in 1'. Create your own pod today at https://podcast.wondercraft.ai/

Hey, thanks for commenting! Totally get what you mean about wanting that authentic connection. That said, more and more people are turning to podcasts for practical reasons (like getting a rundown of the news) rather than just for conversation (like on the Joe Rogan Experience). And with the advancements in AI tech, synthetic voices can sound pretty close to human ones, especially in news readouts. So what we're trying to do here is to offer a useful option for people who prefer to consume information in audio format. It seems that many people found this valuable already, judging by the stats over the past 12 hours.

Hey! Thanks for the nice words!

You're correct about the intonation, and thank you for pointing out that example. Before we run the text through speech synthesis, we try to adjust it to be suited to conversational language. While this helps to address most intonation issues, there are still a few that slip through.

Another challenge we face is using domain-specific language, like C#. It's a bigger issue, but we're actively exploring different solutions to overcome it.

Glad you're finding this to be a useful service, we will be covering a lot more than HN.

Thanks! For text-to-speech we are using models that we have trained on top of the elevenlabs engine.

Hey Piotr - just wanted to say congratz for the awesome work so far man. The quality is genuinely unbelievable. I don't know if you guys are ready to take clients at scale, but I don't see any reason why all newsletter creators wouldn't use your tech right now to address whole new markets. I'll be following the journey, excited for what's to come.