My experiments with creating a digital avatar
From photos of my office to a convincing digital version of me: the images, voice clones, local failures and paid video tests that led me to Aurora.
The full story
I’m a marketer who builds things. These are the projects, experiments and wrong turns that taught me something useful.
By Tyler Casey
From photos of my office to a convincing digital version of me: the images, voice clones, local failures and paid video tests that led me to Aurora.
The full storyThe seven-minute research run, source-checking mistakes and broken restore that shaped how I am building a traceable cannabis research corpus.
The full storyI tested the code, then found the wrong version, a weak demo and a confusing checkout. What my own product releases taught me about finishing the whole journey.
The full storyWhat happened when my 8 GB Mac ran out of space: Fedora builds, a service migration, an unused VM and a storage guard that leaves live work alone.
The full storyLocal search for my Obsidian notes, the stale instructions it uncovered, and the work that made previous decisions useful to me and my coding agents.
The full storyFracture Recordings stopped in 2016. I gave the label a digital archive and my Boski tracks and mixes a radio station, with a few very ordinary web problems along the way.
The full storySwarm Signal and MCPH share a process, but serve very different readers. The broken audio handoff, stale pages and writing fixes that taught me where to draw the line.
The full storyA separate home for my work photos and videos, and the awkward bits of building it: empty imports, missing albums, broken source files and a database recovery.
The full storyNavidrome for my music, Audiobookshelf for listening and Calibre-Web Automated for ebooks. The setup, the messy metadata and the mistakes that shaped it.
The full storyThe details behind the listings, from free-plan limits to the full cost of a subscription.
The coding tools around the models, how access is paid for, and what a stuck ZCode task taught me about where an AI workflow can fail.
Full guideBolt can import an existing repository; Lovable starts its GitHub connection with an export. Code editing, hosting and the practical differences.
Full guideThe difference between Fathom’s recordings and Granola’s enhanced notes, including bot-free capture and the free plans’ history limits.
Full guideAn included software benefit can replace a bill, or just add another login. The difference is in the cost that disappears and the membership behind it.
Full guideA worked explanation of introductory prices, annual payments, credit allowances and the cost of getting something usable out of an AI tool.
Full guideThe break-even calculation for monthly versus annual billing, with a worked example and the assumptions that change the answer.
Full guideThe same model name can sit behind very different products. What a direct subscription, a team seat, a shared login and an API app actually provide.
Full guideElevenLabs’ commercial licence, Synthesia’s avatar and download rules, and Gamma’s exports explain why free access isn’t the whole answer.
Full guideHow I separate personal experience from product research, explain free software and paid plans, and handle old listings and corrections.
Full guideWhy a free building allowance can run out before the month ends, and how temporary grants affect the cost of keeping an app running.
Full guideHow AWS, Google Cloud and Notion distinguish startup applicants, plus the difference between cloud credit and a bill that disappears.
Full guide