How to Make YouTube Shorts With AI (Free Workflow)

My first Short took me six hours. Six. Hours. For forty seconds of content. The second one took forty minutes, and honestly that’s still too long. The difference? I stopped treating Shorts like mini YouTube videos and started treating them like a factory line. Here’s the free workflow I landed on, and it’s the same one I’d hand to anyone starting today.

What We Are Trying to Do

We’re producing a 30–45 second YouTube Short — hook, voiceover, captions, cuts — in under an hour, using only free tools. The brutal truth about Shorts is that volume beats perfection. The algorithm doesn’t care how pretty your edit is; it cares whether people watch to the end. So this workflow is built for repeatability, not polish.

The Best Free Stack for Shorts

Three tools, that’s it:

• ChatGPT (free) — hooks and scripts, over and over
• ElevenLabs (free tier) — the voiceover
• CapCut (free) — auto-cuts, auto-captions, music

No stock footage hunting, no manual captions, no fancy transitions. Shorts don’t need them. If you’re short on video ideas, this free list of 50 faceless YouTube ideas has plenty of Shorts-friendly ones.

What You Need to Get Started

• Free ChatGPT account (or Claude — either works)
• Free ElevenLabs account (~10 minutes of voice a month, which is a lot of Shorts)
• CapCut on your phone or desktop
• One topic you can riff on for 10 videos in a row

How to Do It, Step by Step

Step 1 — Generate the hook with ChatGPT. This is where Shorts live or die, so I batch it. Here’s the prompt I use:

“Write 10 hooks for 30-second YouTube Shorts about [topic]. Each hook must be under 12 words, start with a bold or surprising claim, and create a curiosity gap that forces a watch. After each hook, write the 3-sentence script that delivers on it. Format: Hook — then script. Example: ‘Nobody tells you this about [topic]’ — then a 3-sentence payoff.”

You’ll get ten mini-scripts. What you’ll see: maybe six are actually good, two are weird, and two are ChatGPT pretending to be edgy. Pick the good six, rewrite the delivery line in your own words. This is the whole secret of “AI content” that doesn’t feel like AI content — you filter, don’t create.

Step 2 — Voice it with ElevenLabs. Paste each 3-sentence script into Text to Speech, pick a voice with some energy (for Shorts, boring narration kills retention), generate. If the timing feels off, add a comma or ellipsis in the script — that’s my cheap trick for making the AI pause where a human would.

Step 3 — Auto-cut visuals in CapCut. Here’s the shortcut most people miss: import one longer clip or a short screen recording, then use CapCut’s auto-cut feature with the voiceover as the guide. CapCut will chop the clip to match your narration. For talking-head-style Shorts, this alone turns a 5-minute recording into a clean 35-second video.

Step 4 — Auto-captions. One tap. CapCut’s auto captions are free, fast, and scary accurate. Pick a bold caption style — big text, maybe a highlight color — because 80% of Shorts viewers watch with sound off. Fix any wrong words manually; there are usually one or two.

Step 5 — Music and export. Drop a free track from CapCut’s audio library, keep it quiet under the voice, export at 9:16. Done.

What the Result Actually Looks Like

With this workflow, I can ship a Short in 35–45 minutes from idea to published. The auto-captions come out 95% accurate — I fix maybe one word per video. The ElevenLabs voice gets “what AI tool is this?” comments, which is both a compliment and a warning. One Short in this batch did 41,000 views with zero promotion, just a strong hook and clean captions. Most did a few hundred. That spread is normal — the workflow’s job is making sure the next one exists, not guaranteeing virality.

Where the Free Version Falls Short

Okay, the honest part. ElevenLabs free gives you roughly 10 minutes of voice a month — that’s about 15–20 Shorts, so it’s actually workable, but the free voice options are limited and a couple of them sound a bit “AI-clean.” CapCut’s free version locks some AI transition effects and higher export quality behind Pro. And the auto-cut feature gets confused on fast-paced clips, sometimes keeping awkward pauses. You’ll also feel the sameness: if every Short has the same voice and caption style, viewers eventually notice.

When It Makes Sense to Upgrade

Upgrade CapCut when you’re posting daily and the locked effects or export quality actually matter. Upgrade ElevenLabs when you’re doing more than ~15 Shorts a month, or when you want voices that don’t sound like every other AI Short on the platform. And if Shorts start driving real traffic, treat that as the signal to build a system, not just improvise.

Good Alternatives If You Want to Compare

• Canva Shorts templates — great for text-based Shorts, weaker for voiceover-driven ones
• OpusClip — clips long videos into Shorts automatically, but the free tier watermarks and the cuts need heavy review
• CapCut’s own AI templates — one-tap template videos, fine for testing, forgettable for building an audience

More Free Resources

The free library has the full collection, but start with the free TikTok hooks sample — the hooks work for Shorts verbatim, and they’ll save you the “10 bad hooks for every good one” grind. Then pair it with 50 faceless YouTube ideas when you need topics.

The Paid Systems Worth Knowing About

When you’re past experimenting and want the edits to look genuinely pro, CapCut Mastery Pro 2026 ($69.99) covers the editing side far deeper than any free tutorial. And for the voice side, 60 AI Voiceover Scripts for TikTok, Shorts & Reels ($9.99) is a cheap way to stop writing scripts from scratch entirely.

Try the Voice That Makes Shorts Work

The voice is the difference between a Short people scroll past and one they finish. Grab your free ElevenLabs account here, drop one of your hooks in, and listen to it with fresh ears — if it makes you stop scrolling, it’ll make someone else stop too.