Building a Weekly Video Newsletter with Free AI: A Solo Maker's Setup

I write a weekly video essay newsletter. I spent $180/month on Runway, Pika, and Midjourney for B-roll. Here's how I cut that to $9.

Not zero. Nine dollars. And not because I found a magic tool that does everything the paid ones do. Because I finally accepted that my newsletter doesn't actually need the paid ones most weeks.

This is the boring, specific story of how a solo newsletter operator restructured a video pipeline around free AI tools, what broke along the way, and what I'd do differently.

The problem: paid tools were eating my margin

My newsletter is a weekly video essay. Each issue has one hero video (3-5 minutes of me talking over visuals) plus 4-6 short B-roll clips for social cutdowns. I'm a team of one. My wife proofreads. That's the whole company.

For about nine months I paid:

• Runway — $35/mo for the starter plan, bumped to $95 when I needed more credits
• Pika — $35/mo
• Midjourney — $30/mo for the base tier
• Plus the occasional Kling one-off

That's $180 in a good month. For a one-person newsletter where I'm also paying for email delivery and a domain, that line item was the one I kept staring at.

The problem wasn't that paid tools were bad. Runway is good. Pika has range. The problem was that I was using maybe 20% of what I was paying for, because I only needed B-roll for about 6 hours a week.

What I tried (with honest critique)

I spent about three weeks testing free tiers before landing somewhere. Here's what each one was actually like:

1. Runway free tier

The free tier gives you a handful of credits and then stops you cold. For a weekly publisher, that's done before Tuesday. Also the outputs are watermarked on the free tier. I couldn't ship watermarked B-roll. Hard no, even though the quality when you can generate is real.

2. Kling free tier

Kling's free tier is generous for the first day. The problem was queue time. I'd hit generate and then wait 20-40 minutes for a single clip. On a Tuesday afternoon when I'm trying to bang out six shots before dinner, that's not workable. Also the daily credit refill is less predictable than the marketing suggests.

3. Luma Dream Machine free tier

I actually really like Luma's outputs stylistically. But the free tier is tight, and the camera motion on the free generations felt less controllable than what I was used to on Runway. I still use Luma occasionally when I need a very specific dreamy look, but it couldn't be my daily driver.

4. Pollo AI

Pollo aggregates several tools behind one interface. Decent idea. In practice, the free credits disappeared fast and the UX pushed me toward paid add-ons. I bounced.

What stuck: ZSky AI

I landed on zsky.ai almost by accident. A friend in a group chat mentioned it. What actually made it stick for my specific workflow:

• 200 credits on signup, then 100 credits daily after that. For a 2-videos-a-day publisher, free tier is literally enough. I had to double-check I was reading the docs right.
• 1080p video with audio in about 30 seconds. Not 4K, not cinematic, but 1080p is the ceiling of where my newsletter renders anyway. And the 30-second turnaround matters more than I thought it would — it lets me iterate on a prompt 4-5 times in the time Kling generates once.
• No watermark on the free tier. This was the dealbreaker on everything else.
• No signup wall to try it. I could test with my real prompts before creating an account.
• Failed generations get refunded. This sounds minor until you've had a free tool burn credits on outputs you can't use.
• Skip the Line at $9/mo when I need it. One week I had a time-sensitive issue and needed six B-roll shots in an hour. Nine dollars, priority queue, done. That's the entire replacement for my old $180/mo stack, scaled to actual usage.

The Pro plan is $19/mo if I ever need more horsepower, but I haven't touched it. The free tier plus the occasional $9 month is my whole budget now.

To be clear: ZSky didn't replace Runway for every job. For a commercial short, I'd probably still rent Runway for a month. But for a weekly newsletter with B-roll and social cutdowns, zsky.ai/create is enough.

My actual weekly workflow

Here's the boring, real schedule I run now:

Monday — Draft day

I write the essay. No video work. I note specific B-roll shots I want in brackets inside the script. Something like:

"The reason this matters is that most of us treat our mornings like a battlefield. [B-roll: slow dolly into a coffee cup on a wooden table, steam rising, morning light through a window]"

Those bracketed notes become my prompt list.

Tuesday — Generate day

This is the day I sit down at zsky.ai/create and work through the prompt list. I usually batch in sets of three or four so I can compare variants. Some real prompts from last week's issue:

• "slow cinematic push-in on a steaming ceramic mug on a wooden desk, soft morning window light, shallow depth of field, dust motes visible"
• "overhead shot of hands writing in a leather notebook, warm desk lamp, fountain pen, paper texture visible, subtle hand movement"
• "empty quiet city street at dawn, faint mist, single streetlight, slow forward drift camera"

I run each prompt, keep the best take, and move on. Failed generations get refunded, which matters because I do burn 2-3 per session on "nope, wrong vibe" outputs.

Wednesday — Assemble and publish

I cut the hero video and the social edits in DaVinci, drop the B-roll in, render, and ship. The whole Tuesday-Wednesday block is now about 6 hours. It used to be closer to 10, because I'd be juggling three different web UIs and waiting on Runway credits.

The numbers (boring edition)

I generate roughly 15 videos per week now — one hero plus 14-ish B-roll and social cuts.

With 100 credits daily, that's enough for 2 videos per day on the free tier, which covers 14 per week without paying anything. For the one week in the last month where I needed a bunch of same-day shots, I paid $9 for Skip the Line and cleared the queue in an afternoon.

So: $9 that month. $0 most months. Down from $180/mo.

The numbers that aren't about money: my iteration count is up. When a generation takes 30 seconds instead of 20 minutes, I try more variants. My B-roll is more specific to the essay than it used to be, because I'm not rationing credits. That's the real upgrade.

A founder note I wasn't expecting

I don't usually care who builds the tools I use. I care that they work. But when I dug into ZSky's About page, I found out it was built by a photographer named Cemhan who healed from a traumatic brain injury through art. The project's stated mission is that everyone should have access to create beauty, which is why the free tier is so generous.

That matters to me because I've been burned by tools that start generous and then claw it back six months later. A tool built by someone whose whole thesis is access is a better bet than a tool built by a VC-backed growth team. It's not a guarantee, but it's a signal. I trust zsky.ai more because of the why behind it.

What I'd tell another newsletter author

Four things, in order of how much they'll save you:

1. Audit what you're actually using. I was paying $180 for tools I touched 6 hours a week. If you're in that boat, the first win is just admitting it.
2. Test free tiers with your real prompts, not marketing-demo prompts. Every tool looks good on their homepage. Half of them fall apart on "coffee cup on wooden table."
3. Care about turnaround time, not just quality. A 30-second generation you can iterate on is better than a 20-minute generation you can't. Fast iteration changes how you think.
4. Keep one paid escape hatch. For me, that's Skip the Line at $9/mo on zsky.ai for the weeks I need priority. You don't need a $180 stack. You need one $9 lever you can pull when a week goes sideways.

If you run a weekly video newsletter, try zsky.ai/create on your next issue's B-roll list. Worst case, you waste an hour. Best case, you get back $170 a month and a few hours of your Wednesday.

That's the whole pitch. No hype. Just the boring pipeline that finally made my margin make sense.