I built my own AI video editor for my agency videos

I'm building my own AI video editor and it's been a huge eye-opener for models and the productivity you can gain from AI. The image below left is pre edited, post edited. The idea is I want to record informational videos around SEO/Link Building and AI to share good ideas, get YouTube mentions for AI search and just build a real presence for the brand and myself. The process is I record a video on my phone, slap it in iCloud Drive and have my iMac (my Mac mini equivalent) turn it into something I'd actually want to post. Captions, visuals, thumbnail, the lot. Then I can watch it, approve it and it automatically books it for posting on all platforms. < this all works and is amazing! Getting the edit right has been the painful bit. The visual quality choices have been hit and miss, some it would absolutely smash to perfection, others would come back pretty terrible. Also breaths and non-talking scenes have persisted, but Gemini 3.7 Flash has been amazing for this. The real breakthrough was switching from Fable 5.1 to GPT Astra for directing visuals. Fable would miss a lot and try to build really crappy animations when unsure, but Fable has been amazing at making the scenes Astra chooses come to life. The downside? It's not cheap. You need the frontier models to have a chance at a good edit and it needs to be on Max effort (tried lower 200+ iterations) I have two 20x GPTs 2 5x GPTs and currently I get about 6/7 shorts a day and one long-form (17 minutes on average) a week out of the whole weekly sub for all accounts. Edit visual/quality? Decent - really good if you give it a detailed brief (mine it figures out based on the video). It's not like the epitome of edits but you'll get the level of like a decent editor. Price per minute $600 / 255 minutes of video = $2.36 a short! The video was a long form Astra edited, took 5 revision sessions: https://www.youtube.com/watch?v=PrrjQbMxAqA&t=11s Was pretty painful watching a 16 minute video that many times but hopefully that part gets easier, short form is way more accurate and gets the video bang on every time more often.