Timi

@timigod.eth

166 Following

1338 Followers

Timi pfp

There are still some quirks, like cleaning up the audio quality, for example. Here’s what Fable will look like when/if we finish building it.

0 reply

0 recast

0 reaction

Timi pfp

I wanted to explore rebuilding the NotebookLM Podcast feature and making it more accessible. Here’s a snippet of a 30+ minute podcast generated by my app after I uploaded The Little Bitcoin Book to it.

1 reply

0 recast

1 reaction

Timi pfp

Yeah I don’t think it’s super likely without better batteries. I think the short term solution is faster & more efficient local models. https://warpcast.com/timigod.eth/0xd2f0ff92

0 reply

0 recast

1 reaction

Timi pfp

If that’s the case, then nobody will make 1 hour videos on desktop either. 🤷🏾‍♂️

1 reply

0 recast

0 reaction

Timi pfp

Well, that makes my point. Today, people are creating 30-second to 5-minute videos on their phones, a big leap from nothing just a few years ago. In the next 5 years, it could easily become 30-minute or even hour-long videos, especially with AI driving the shift. Almost every task is being done more on mobile: gaming, photo editing, coding, document editing. ALL of these had zero mobile activity not that long ago.

1 reply

0 recast

0 reaction

Timi pfp

*GPU intensive, you mean? I’m pretty sure all mobile video editing tools like iMovie, LumaFusion etc do some video encoding. Things always start out as “being better/only possible on desktop”, but they always eventually make it to mobile.

1 reply

0 recast

0 reaction

Timi pfp

1 reply

0 recast

0 reaction

Timi pfp

But they’ll always have an “unfair”advantage for their own offerings.

0 reply

0 recast

0 reaction

Timi pfp

The solution is probably just MUCH faster local models. Apple has never allowed meaningful background processing and yet, apps do a lot of cool, previously considered time consuming things.

1 reply

0 recast

0 reaction

Timi pfp

I’m playing around with a NotebookLM podcast generator type knock off and even if there was a local tts model that was lightweight enough, the user would still have to wait ~3-5 mins without closing the app for a podcast to be generated.

2 replies

0 recast

0 reaction

Timi pfp

Increasingly capable local models (like Llama 3.2 1b & 3b) make me very optimistic about inference at the edge - mostly because of cost. However, I worry about how limited they’ll be if Apple continues to disallow any kind of meaningful background processing on iOS.

1 reply

0 recast

0 reaction

Timi pfp

I don’t understand Unichain (and a lot of other “app chains”)

0 reply

0 recast

0 reaction

Timi pfp

0 reply

0 recast

1 reaction

Timi pfp

I still can’t get over seeing Linkin Park this week.

0 reply

0 recast

3 reactions

Memecaster pfp

@memecaster.eth

You can now add your favorite Farcaster accounts and allocate a daily Degen tip allowance to them. Start tipping and spread the love! 🎩 Try it out on the app (Links in reply)

5 replies

6 recasts

5 reactions

Timi pfp

Just rented Robot Dreams and found out that it’s entirely dialogue free.

1 reply

0 recast

1 reaction

Timi pfp

With time, the differences/advantages can increase. FWIW, I built Quick Snack entirely with Cursor. I love Cursor.

1 reply

0 recast

0 reaction

Timi pfp

I guess the primary advantage is that you don’t have to set up anything to have something tangible in your hands that you can actually use. Setting up Expo etc w/ cursor is a little long, even when you’re not a noob, it should be abstracted.

1 reply

0 recast

0 reaction

Timi pfp

1 reply

0 recast

1 reaction

Timi pfp

I’ve had a lot of fun tinkering with this over the past couple of weeks. With https://quicksnack.dev, you can rapidly build React Native apps using AI. See demo (1.2x speed):

3 replies

3 recasts

5 reactions