That's something I haven't seen in a while!

Comments URL: https://news.ycombinator.com/item?id=47363480

Points: 2

# Comments: 0

We built Lucent Chat because creating with AI has become a mess. You'd ask ChatGPT for a prompt, copy it into another tool, tweak settings, open three tabs, run into token limits, and lose your style halfway through. Iteration was slow, context vanished, and the process killed the creative flow.

Lucent Chat is our attempt to fix that - a single chat where you can generate, refine, and manage all your visuals and videos across Veo, Sora, Kling, Seedream, Nano Banana, and more. You describe what you want and Lucent handles the rest. It will find the right model, optimise your prompt + parameters behind the scenes.

Once you do generate something, it's dead simple to keep on going and iterate off it with remixing. You can also get inspiration from what others are creating.

We started this out of frustration with how fragmented creative AI had become, and now we're releasing it to the world.

We're giving out free credits (40 images or 4 videos with Sora 2) to early adopters from HN!

You can try it here: https://lucentchat.com?utm_source=hn

Feedback is hugely appreciated.

Comments URL: https://news.ycombinator.com/item?id=45750799

Points: 10

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=44646021

Points: 7

# Comments: 0

For us it was a bit different because we were doing robotic F&B so people could come to our restaurant to see the place from Reddit IRL.

I would target memes in niche dev communities if you want direct outreach. Like if you're looking for people to test your product you'd have more luck on a 100 upvote post in some obscure deeptech subreddit then you would if your post got 10k upvotes on r/funny.

Sorry for any grammar mistakes - writing this on the rush hour train :)

Comments URL: https://news.ycombinator.com/item?id=44129994

Points: 5

# Comments: 0

Yup. Must have been a horrific wake up call :(

Really well written and insightful.

I'm in discussions with friends of Electro on what approach would be best to take. I go into more detail on this announcement made on the discussions board: https://github.com/pTinosq/Electro/discussions/17

I will take your advice into account for future updates. Thank you

I had a meeting today with 6 other software engineers who are interested in the project & have experience in computer graphics (a topic I will admit is not my strongest) and we all unanimously agreed to gradually shift Electro away from WebView2 and towards a custom-built 2D renderer for truly native "hyperfast" performance :)

As mentioned in another reply on this thread, I'm not here to fool anyone - this is version ~4.0 of Electro and the first version I feel is feature rich enough to ship publicly. The other 3 versions of experimentation (over the last year and a half) included Rust, C++ & OpenGL, Electron, and Tauri all of which had significant pitfalls.

The final decision was to "just ship it" and see how the community reacts. I've seen (as I anticipated) a huge number of people suggesting we move away from Tauri and that's exactly what we will do.

Open sourcing this project was keeping me up at night because deep inside I knew that it would need another version. That being said, I'm so glad I did it or else there would not be any pressure to actually do it and it probably would have ended up just like all my other dozens of projects that are just sat in private repos collecting dust. At least now with this version people can A) Enjoy using a new tool w/ unique features (built-in terminal, performance focus) and B) Even contribute if they feel like doing so.

I'm very excited to see where this goes but the compass is very clearly pointing towards ultra low-level rendering techniques and I honestly can't wait to set sail!!

He also kindly pointed to this GH Issue which discusses your privacy concern: https://github.com/MicrosoftEdge/WebView2Feedback/issues/105...

This quote from the issue stood out in particular: " WebView2 is considered a Windows component, and the data collection consent is governed by Windows Diagnostic setting on Windows 10 as a centralized switch.

End users are empowered to control the data collection of WebView2 and can do so via toggling the Windows Diagnostic setting on Windows 10. This is also what the Edge browser does. On Windows 7/8.1, because there is no Windows Diagnostic setting, we treat this as no consent for optional data. There is very limited required data that the OS always collects, unless you're on some specific SKUs. Developers are definitely welcomed to convey that to their end users and ask them to use the OS toggle. "

I've been meaning to move away from Tauri + WebView2, this might be the best call to make (not only for this reason of course)

As soon as I can port over core functionality to a custom desktop renderer things will go from being 10x faster (currently) to 1000x faster (the end goal).

I'm currently trying to balance Electro with: university final year, job applications, 2 other OSS projects I contribute to, and a friends' business website so these features will most likely take many months to get through. I'm hoping the eyes Electro has gained from this announcement will bring in some talent who know more about lower level image viewer functionality than I do so that they can help guide the project towards the ultimate performance-first tech stack!

I downloaded NASA's photos of the moon and tested those with wonderful results. I was actually using these enormous 16k+ images during my testing specifically because they are the ultimate benchmark for image viewer performance!

Tauri 2.0 was initially chosen because it would let me get an MVP out quickly to start getting user feedback (like your own). My end goal is to move to a custom renderer so I'm not relying on Chromium / WebView2. This will take many months of work I suspect (balancing with my FYP @ university & other projects).

Thanks for bringing Oculante to my attention - it looks awesome! That level of functionality is what I'm striving for (w/o as much UI clutter though).

I feel the TS & CSS pain. Electro v1.0 was C# WPF and very very very fast but the UI was just rubbish and I knew C# WPF would hold me back in the future. Electro v2.0 was then written in C++ but I didn't get far with it because I REALLY didn't like writing C++. Electro v3.0 was then written with Electron + TS + CSS but electron is shit. Finally, I landed where we are today - it struck the perfect balance of performant, modern language, fun to write, allowing for rapid iteration.

Will Electro stay with the TS+CSS+Tauri stack forever? Probably not! As the community grows I hope bring in contributors who are more well versed in the world of low-level image viewers to help guide Electro to be EVEN FASTER & closer to the system level (Like Electro v2.0 aimed to be w/ c++).

For the time being I've decided to just ship with what I have and iterations will come later. I do dream of the day we go frameworkless (to an extent ofc) and render everything from the GPU at nanosecond speeds while also maintaining the sleek feeling Electro currently has!

Electro aims to do what an image viewer should do best - show images. There aren't any UI elements when an image is first loaded - hence the hidden terminal (pack a lot of functionality without worrying about visual clutter).

I have been carefully thinking about the direction I want to take with UI elements. The goal was to build the fastest & most efficient image viewer. Will I achieve this through clickable UI elements? Through keybinds only? Through terminal? Only time and testing will tell.

The project is fresh out of the oven and ready to be moulded into whatever the community wants it to be. The two mantras Electro will always stand by are Performance & Efficiency.

Thanks for the thread - loved reading everyone's thoughts.

P.S JPEGView was by far my favourite but again, too ancient. I want something truly fresh and that was the goal of this project

The project is in its infancy at the moment. Features like keyboard controls to zoom etc. are all on the table - I love when users give me all these ideas for where to take the project