David Millington, now working for RemObjects, will be giving a presentation via Embarcadero’s online channels early in the morning wednesday Perth time.
We won’t be stealing any of that thunder, but we will have him as a guest, along with ADUG President Geoff Smith, to talk in general about their experiences with “closed loop” LLM use .. ie where there is an automated continuity between generating code, compiling, feeding back the output or errors, and then responding to the feedback with more code generation.
I’m sure we will also touch on favourite models/providers, as well as tests, and documentation, and planning, and more.
I will come back with a Zoom link, after I double check the details.
Just so I understand - are you saying that with OpenShot it took five hours to render a ~1 hour 44 minute video - and when it did it created a 7gb file?
You then used ShotCut and the same video took fifty minutes (5 times faster) with a roughly 70% smaller file which, even then, was 2gb.
That is absolutely useless. As you can imagine for my role I render a lot of videos, for example every single one of the 42 CodeRage recent videos. If it took five hours it would be impossible for me to do. Even 50 minutes is a long time - for me it takes roughly 12 minutes for a one-hour video, if that.
To give you some comparison, a video by Brazilian MVP Dion Mai came in at 56 minutes in length - file size on disk after rendering 308mb (including “in vision” subtitles) at 1080p MP4. Adding .SRT subtitles usually only adds a few MB at most. Invision ones don’t add anything since they are rendered as part of the image; however, they can interfere with auto-subtitling by platforms like YouTube and LinkedIn.
YouTube have begun an experiment where they will add alternative audio tracks to presentation videos - in several languages - and sync the speaker’s lips with the AI dubbed audio. It is quite incredible. This happened on one of my Brazilian colleague’s videos and it was very weird to hear him speaking English (it cloned his voice) and it was very accurate, including the same emotions and style of vocal delivery. It’s definitely the future.
Rendering at 2k and 4k is absolutely pointless too - most streaming platforms will down-sample your videos to 1080p. This is especially true for YT, Meta, and LinkedIn, where they serve up down-sampled content for mobile or for when they detect slow connection speeds.
The only time to use 2k or 4k is if you are delivering via a CDN with sufficient bandwidth. Even then I do not recommend it. Our stats show that most people could not sustain that throughput and you would likely end up with people saying the video was choppy or buffering if you’re streaming it. Even for a website video it’s absolutely not recommended to serve up 2k/4k if it’s a site video.
What’s your setup, @ianbarker ? (cpu, gpu, software) I don’t use anything like Adobe Premiere or the like.
This video was mainly just webcams, but when we have code on screen or showing web pages (most times) it’s better in 4K.
And YouTube lets you choose whatever in the resolution department, for playback.
If I wasn’t adding the little intro bit, then we’d just cut the relevant 4K section from Zoom (immediate) and upload to YT.
The 5 hour thing is running on just CPU seemingly .. although I’ve never come across that kind of duration before.
(Actually it was 6 hours first time, because I accidentally chose 30 fps instead of the 25 fps of the original file)
The key thing is the GPU (although that is pretty old now too). It’s a NVIDIA Geoforce RTX 2070 Super.
My machine has 4k, 2k, and 1920 x 1080 screens attached. The 180p screen is the one on the left and that’s the one I share for the webinars since it will definitely be at the right resolution.
The Mac is to my right with a 4K TV as the monitor. It spends most of the day playing weird music or showing calendars (I have a lot of meetings).
I use NVIDIA Broadcast when I am doing webinars - this provides studio effects like echo cancelation, AI noise reduction, and so on. It is so good I can literally play guitar while I speak and you would not know!
The mic is a regular USB mic - but it’s sitting in an acoustic shield. The NVIDIA broadcast does the rest.
I use SSDs - exclusively - and got the fastest ones available. The write speed makes a lot of difference for the rendering. I use Camtasia, latest version. I have tried other things, but Camtasia is easily the best. If I have to convert video from one format to another I use Duo video convertor.
Perth Meeting - Online - Tuesday 10th February 2026
Hello all.
Perth ADUG Meeting at 6pm tonight.
Some general chat, and a bevy of news items from LinkedIn and fb that have appeared.
The ADUG Symposium 2026 will be on Friday 8th May … with 4 speakers in-person at the Melbourne venue. (Live and online attendance). And a special extra online session Saturday with a highly admired Delphi Developer and Author.
