dgerard

https://www.youtube.com/watch?v=hYZzOZSe6_0 - video

https://www.youtube.com/watch?v=HBL_tD_x3OA - video, which came out very well

I wrote this and I want to know what else someone would want to know, if I'm dead wrong on any of it, etc.

the channel: https://www.youtube.com/@pivottoai

How I do the Pivot to AI YouTube videos

Cat: site news
Img: a title card, or me looking horrified at s/t

The Pivot to AI video series is getting views and subscribers, which is nice!

It’s a couple of hours to write a post and expand it into a script, fifteen minutes to record if nothing breaks or falls over, and another hour or two to clean up the audio and assemble the video, including screenshots of sources.

The philosophy is: your content is what matters, everything else is a bonus. Put in effort, not money. We’re making punk rock here. I did fanzines in the ’80s and books in the 2010s on the same principles.

Fortunately, cheap consumer electronics is good enough in 2025. Here’s how I do videos on a budget of zero.

-- read more --

Camera: I use my phone, which is the best camera in the house. It does 1080p on the front-facing camera. This records H.264 video and 96kpbs AAC sound, which is fine for voice.

I try to do everything in a single take. Fancy is my enemy. “We’ll fix it in post” is a film-maker phrase meaning “well, that was a cock-up.” Everything you fix in post takes ten to twenty times longer than getting it right the first time.

Camera mount: anything that will mount a phone stably. The loved one and kid got me a ring light for Christmas and told me to rant on TikTok when something annoyed me. Unfortunately, ring lights are incompatible with wearing glasses - you get virtual googly eyes projected perfectly onto your lenses - but the phone holder bit still works well mounted on the desk. I also have an older phone mount for a camera tripod.

Microphone: Your video can be iffy, but your sound has to be good.

I use my Jabra Evolve 40 headset, which was designed for work Zoom calls. This is not great, but it’ll do. I turn the bass up in post-production and it sounds better.

I really want a Røde - the loved one has a Røde M3, which is all the mic you need for a remarkable range of use cases, for around £80. The M3 really needs phantom power - if you use a battery, you will forget it’s switched on and it’ll go flat - which is another £20 box. But Røde make an enormous variety of good podcasting mics of professional quality and you’re unlikely to go wrong. A Røde is the next piece of kit I’ll spend money on.

Lighting: A 9W daylight bulb above, a 9W daylight bulb to my left, and a 9W yellow bulb (“warm white”) to my right are doing the job so far. The lights to my left and right are in clip-on gooseneck mounts attached to the shelves.

Teleprompter: Elegant Teleprompter on Android. This puts a floating window over your camera app. It’s free with unobtrusive ads and the developer is very in-touch with his user base and likes to fix problems.

Sound: I edit the sound in Audacity, which is free, open source, and just works. If you know what you want to do, you can probably do it. I normalise the very quiet phone audio, do a noise reduction, pump up the bass 4.5dB, then go through the audio de-umming and removing breath noise, the latter being the main reason I want a microphone that isn’t up my nose. Then compression, then it’s ready.

Screen shots: I take these in Firefox and edit them in GIMP. I also make the title cards in GIMP. Just make things 1280×720.

Video editing: I use OpenShot, which is a bit open-source, but it basically works and it’s free. I get the raw video, the cleaned audio, the various still images, and the theme music, and assemble the final video. Export at 720p as “MP4 (H.264 va).” I could go to 1080p, but this is a talking head show and you don’t need my nose hairs that sharp.

ffmpeg: this is the Swiss army knife of video and about as fiddly to use. I’m very into tweaking things in ffmpeg.

OpenShot had trouble with the Nature video, which kept freezing in rendering the 7-minute raw file late in the video. I ran the raw video through ffmpeg to add a key frame every ten frames, which is probably overkill, but it worked, so I’ve kept doing it. I do not recommend you do this unless you specifically have this problem. But, to generate an oversized video with no sound (because the cleaned-up audio track is separate) and a ton of extra key frames:

ffmpeg -i input.mp4 -vcodec libx264 -x264-params keyint=10:scenecut=0 -crf 14 -an output.mp4

I can also do things like stretch a clip to 1.2× length to make speech clearer:

ffmpeg -i fatima.mp4 -vf "setpts=1.2*PTS" -af "atempo=0.8333" fatima2.mp4

When you finally give in to the urge for fancy with equipment, pro podcaster kit is super cheap on Alibaba and hence Amazon. Go through surplus listings and see if you’re lucky.

https://youtu.be/eEgmF7x-G5o - video

https://www.youtube.com/watch?v=dI1ZLKM2t0k - video

originally posted by me to tumblr in 2016, but still completely accurate

https://www.youtube.com/watch?v=oqHi3806cSc - video

While this linear model's overall predictive accuracy barely outperformed random guessing,

I was tempted to write this up for Pivot but fuck giving that blog any sort of publicity.

the rest of the site is a stupendous assortment of a very small field of focus that made this ideal for sneerclub and not just techtakes