The Solo Studio Stack: What I Pay For and What I Self-Host

I run a one-person creative studio and my monthly tool bill sits around 180 EUR. That number sounds small until you realize it replaces roughly two full workdays a week. Here is exactly where I pay and where I self-host, plus the three tools I would never cancel.

The Rule I Use To Decide Pay Versus Self-Host

Every tool decision comes down to one question: what happens when it breaks at 2am? If a broken tool means a client sees a blank page, a scheduled post never fires, or a checkout fails, I pay for the managed version. If a broken tool just means I rerun a batch job in the morning, I self-host it.

That single rule has saved me more grief than any pricing spreadsheet. I used to self-host everything because I liked the control and the low cost. Then I spent a Sunday afternoon debugging a database that had silently corrupted, and I lost a week of scheduled content. The 15 EUR a month I saved cost me maybe 12 hours of recovery plus the reputation hit of going dark.

So now I split my stack into two buckets. Bucket one is customer-facing and time-sensitive: the storefront, the scheduler, the payment layer. Those get real subscriptions with real support teams behind them. Bucket two is internal batch work: image processing, file conversion, backups, log parsing. That runs on a cheap box in the corner of my apartment and a small cloud instance.

The math is simple once you stop thinking about the sticker price. A 20 EUR subscription that saves me three hours a month is paying me back at a rate I would never turn down for freelance work. A self-hosted tool that saves me the same 20 EUR but costs me two hours of patching every quarter is a worse deal, even though it looks cheaper on the invoice.

I check this line every quarter. Tools move. A service that was worth paying for last year sometimes gets a self-host option good enough to switch. And a self-hosted setup I loved sometimes becomes a maintenance sink I quietly resent. The rule stays the same even when the answers change. If you want the full picture of how I structure the whole operation, the Claude Blueprint walks through the system end to end.

What I Pay For And Why It Earns Its Slot

The storefront is non-negotiable. I run on Shopify and I have zero interest in self-hosting a checkout. Payment infrastructure is the one place where a bug does not cost me time, it costs me trust and possibly the whole sale. When someone hits buy, that transaction has to clear the first time. I pay for that certainty and I never think about it again. No patching, no PCI compliance headaches, no 2am pages when a card processor changes an API.

Social scheduling is my second paid slot. I use Buffer to line up a week of posts in one sitting. I tried self-hosting a scheduler through a cron job and a few API calls, and it worked until an access token expired silently and three days of posts never went out. That gap cost me the momentum I had built over a month. The managed scheduler handles token refresh, retries, and platform quirks that change without warning. Ten minutes of setup a week versus an ongoing debugging project. Easy call.

Voice generation is my third. I use ElevenLabs for narration on video content. There are open models I could run locally, and I have tested a few. The quality gap is real but shrinking. What keeps me paying is the speed. I generate a two-minute script into finished audio in under 30 seconds without spinning up a GPU or babysitting a model. When I am producing five pieces a day, that latency difference compounds into hours.

The pattern across all three is the same. I pay when the tool sits on the critical path of something a customer sees or something that has to happen on a schedule I cannot babysit. I pay when the failure mode is invisible until it is too late. These are not luxury purchases. Each one replaced a self-hosted attempt that broke in a way that cost me more than the yearly subscription. The invoice looks like an expense. It behaves like insurance against my own worst afternoons.

What I Self-Host And What It Actually Costs Me

Batch image work runs on my own hardware. I process, resize, and organize thousands of generated images a month, and running that locally costs me nothing beyond electricity. When a batch fails, I see it in the morning log and rerun it. Nobody is waiting on it. The stakes are low, the volume is high, and the job is predictable. That is the exact profile for self-hosting.

My upscaling flow is a hybrid. For high-volume rough passes I run local models. For the handful of images that go on product pages or thumbnails, I send them through Magnific because the detail it recovers on the final hero shots is worth the per-image spend. This is where the pay-versus-host line gets interesting. It is not always a whole category. Sometimes it is a split inside one workflow: self-host the 95 percent that is throwaway, pay for the 5 percent that ships.

Backups are self-hosted and I check them weekly. File conversion, log parsing, metadata tagging, and asset renaming all run on scripts I wrote once and now forget about. The maintenance cost is real but small: maybe two hours a month across everything, usually when an OS update breaks a dependency. I budget for that. It is a known cost, not a surprise one.

The trap with self-hosting is underestimating the ongoing tax. A tool is not free just because there is no monthly charge. Every self-hosted service is a thing I have to update, secure, and eventually debug. I have killed three self-hosted tools in the last year because the maintenance quietly grew past the value. One was a note database that needed a full migration every time the app updated. Another was a media server that ate a weekend when a storage drive filled up silently.

My honest rule for self-hosting: it only wins if I can go a full month without touching it. If a tool needs my attention more than once a month and it is not customer-facing, it goes on a subscription or gets replaced. Attention is the real currency in a one-person studio, not cash. I have plenty of the second and almost none of the first. Every hour I spend patching a box is an hour I am not producing the work that actually pays the bills.

The Three I Would Never Cancel

If I had to cut my stack to three tools tomorrow, here is what survives and why.

First, the storefront. Without a place to sell, nothing else matters. Shopify is the floor of the whole operation. It handles the checkout, the tax logic, and the order flow that I have no desire to rebuild. Cutting this would mean cutting the business itself.

Second, the AI writing and reasoning layer I build everything else around. This is the tool that drafts, edits, and structures nearly every piece of text I produce, including the first pass of this article. It is the closest thing I have to a second employee. It reads my briefs, follows my rules, and produces work I can refine instead of write from scratch. The use here is not incremental. It is the difference between shipping five things a week and shipping one. I covered how I wired this into my daily flow in the Claude Blueprint.

Third, the scheduler. Buffer keeps my presence consistent even on days I do not open the laptop. Consistency is the thing that actually builds an audience, and a one-person studio cannot be consistent through willpower alone. I need a system that fires whether I show up or not. This tool is that system.

Notice what did not make the list. Voice, upscaling, and all the batch tools are valuable, but each has a workable fallback. I could narrate in my own voice, upscale locally, and process images by hand in a pinch. Painful, but survivable. The three above have no cheap fallback. Losing any one of them would break something that keeps the studio running.

That is the real test for a keeper tool. Not how much it costs or how clever it is, but whether the studio still functions without it. If the answer is no, it earns a permanent slot no matter the price. Everything else is negotiable, tradeable, and worth reviewing every quarter when the pay-versus-host line inevitably shifts again.

Bottom Line

My whole stack costs around 180 EUR a month and buys back roughly two workdays a week. The split that makes it work is simple: pay for anything on the critical path where silent failure costs trust or momentum, and self-host the predictable batch work that can wait until morning. The subscription price is not the expense to watch. The attention tax on self-hosting is. In a one-person studio, my hours are the scarce resource, not my cash.

If you are building your own setup, start by sorting every tool into customer-facing versus internal batch, then ask what breaks at 2am. That one question will make most of your decisions for you. The three tools I would never cancel each pass a harder test: the studio stops working without them. Everything else stays under review. If you want the full system these tools plug into, the Claude Blueprint lays out how the pieces fit together and where the use actually lives.

This article contains affiliate links. If you sign up through them, I may earn a small commission at no extra cost to you. (Ad)