Search anything about video automation and the word "free" is attached to it: free subtitle generator, free JSON to video API, free AI video generator API, free TTS. It comes back on almost every page of this site, which is a good sign that nobody has answered it properly.
So here it is. Not "our plan is better", but what free actually buys you in each layer, and where the bill starts.
Free always costs something. The question is what.
Generation and rendering burn GPU and CPU time that someone pays for. That means a free tier is never free of consequence — it is paid for in one of five ways:
- A watermark on the output
- A queue, so your job runs when paying customers are done
- A cap: minutes per month, characters per month, renders per month
- A resolution or length ceiling
- Your content, used to improve the service
When you compare free tiers, compare which of those five you are agreeing to. That is the actual difference between them, not the feature list.
Captions and subtitles
This is the one place where free genuinely wins for most people. TikTok, Instagram, Snapchat and YouTube all write captions from speech at no cost, inside the app, for one clip at a time. If you post a handful of videos a week to one platform, stop reading here and use theirs.
Free stops mattering at three specific moments. When you need the transcript as a file you can reuse, because the platform keeps it inside the app. When you need the same caption styling on every video, because the platform gives you its styles, not yours. And when you cut the same clip for four networks, because you will transcribe the same audio four times.
X is the exception that catches people out: it does not transcribe your upload at all. You attach a SubRip file, and it is strict about the format. See the captioning workflow for generating one.
Text to speech
Free TTS tiers are capped by characters per month and come with a limited voice list. That second part is the trap. If your channel has a voice, and the free line-up changes — which it does — your back catalogue and your new videos no longer sound like the same channel.
Budget for a paid voice the moment consistency matters. Until then, free is fine for testing whether narration works at all.
JSON to video and rendering
Rendering is metered by output: per render minute, per render, or per credit. Free tiers here exist to prove your integration works end to end, and they are sized for exactly that. You get enough renders to submit a job, poll it, and see the file come back.
What they are not sized for is production. The moment you render on a schedule, the free tier runs out mid-month and your pipeline stops without telling you. Price your real volume before you build on it — see how the render step works and what it meters.
AI video models
This is where "free unlimited" claims are most common and least true. Generating a few seconds of video is one of the most expensive things you can ask a computer to do. Any service offering it without limit is either putting you behind a very long queue, watermarking the result, capping the resolution to something you cannot publish, or running on credits that quietly expire.
There is one legitimate free move here: most providers give new accounts a starting balance so you can compare models on your own prompt before paying. Use it for that, then decide. The model overview lists what each one costs per job so you can compare without signing up first.
Automation platforms
n8n's Community Edition is free to self-host, and that is a real free, not a trial. You pay in server time and maintenance instead: updates, backups, and the memory it eats when a workflow pulls a large file. For a handful of workflows a small VPS is enough. See the self-hosting guide for what breaks first.
Make and Zapier both have free tiers with a monthly allowance — operations for Make, tasks for Zapier. Both are enough to build and test a scenario end to end. Where they diverge is a polling loop: Make counts every module run as an operation, so a workflow that checks a render every ten seconds burns your allowance on waiting. n8n counts the whole run as one execution, so the same loop costs nothing extra.
How to compare without getting surprised
Three questions settle most of it.
What is the billing unit? Render minutes, generation credits, characters, operations and executions are not comparable to each other. Convert everything to your own monthly volume before you look at a price.
What happens when the free tier runs out? Some services stop, some queue, some silently degrade quality. The ones that stop are the honest ones, because you find out immediately.
What are the output rights? Free tiers often come with different commercial terms than paid ones. If the video is for a client, read that clause before you render.
The short version
Use free for captions if you post to one platform by hand. Use free tiers everywhere else to prove the integration works, then price your real volume. And treat any "unlimited free" claim as a question rather than an offer: unlimited how, paid for by whom?
Build your first automated video
One API key for deterministic JSON-to-video plus AI video & image generation. Documented and ready for your pipeline.