An AI YouTube Shorts generator turns an idea into a finished vertical video by handling the slow parts for you: the script, the voice-over, the B-roll clips and the captions. The workflow that holds up best is to write a hook-first script of about 150–170 words, voice it, generate five or six 9:16 clips that match each line, add burned-in captions, and then tick YouTube's "AI use" box when the result looks realistic. This guide walks through each step with real tools, real credit costs and the YouTube rules that decide whether your Short can earn money.
Updated 20 October 2026. YouTube rules are taken from YouTube Help pages as of that date; Kenerate credit prices come from the app's own model configuration on the same day.
What is an AI YouTube Shorts generator?
The phrase covers two different kinds of tool, and it helps to know which one you need before you start.
- All-in-one "prompt to Short" apps. You type a topic, and the app writes a script, picks stock footage or AI clips, adds a synthetic voice and captions, and exports a 9:16 file. InVideo, Fliki, Pictory and similar tools work this way. They are fast, but every channel using the same template tends to look the same.
- A modular workflow. You use the best tool for each layer: a script you write (or draft with a chatbot and edit), a dedicated voice generator, a video model that makes original clips, and a caption tool. It takes longer, but each Short looks like your channel, which matters for YouTube's monetization rules (more on that below).
This post covers the second approach, because it is the one that survives YouTube's "inauthentic content" policy and the one you can control shot by shot. You can still keep the whole thing inside one account: on Kenerate, the voice, video and subtitle tools all draw from the same credit balance.
What does YouTube count as a Short in 2026?
Before generating anything, get the format right. According to YouTube Help, a video counts as a Short when it is up to three minutes long and has a square or vertical aspect ratio. Wider videos, such as 16:9, stay long-form even if they are short.
| Spec | What YouTube requires or recommends | What to set in your tools |
|---|---|---|
| Length | Up to 3 minutes (uploads after 15 October 2024) | 20–60 s for most faceless formats |
| Aspect ratio | Square or vertical | 9:16 for every clip |
| Resolution | 1080×1920 is the common target for 9:16 | 720p drafts, 1080p for finals |
| Copyright claims | A Short over 1 minute with an active Content ID claim is blocked globally | Use original or licensed audio only |
| AI disclosure | Required for realistic altered or synthetic content | "AI use" set in YouTube Studio at upload |
That copyright row is easy to miss. If you go past one minute, a single claimed music track can block the whole Short worldwide until the claim is resolved. Generated clips with their own audio, or a clean voice-over with no background track, avoid the problem entirely.
How do you make a YouTube Short with AI, step by step?
Here is the full workflow. Each step lists the Kenerate tool and what it costs, so you can see the total before you spend anything.
Step 1: Write a hook-first script (150–170 words for 60 seconds)
Spoken English runs at roughly 150–170 words per minute in a brisk Shorts read, so a 30-second Short is about 75–85 words and a 60-second Short about 150–170. Structure it like this:
- Line 1, the hook (0–2 s): the payoff or a question. "This setting doubles your Wi-Fi speed" beats "Hey guys, welcome back."
- Lines 2–5, the beats: one idea per sentence. Each sentence will become one shot.
- Last line, the payoff or loop: answer the hook, or end on a line that flows back into the first one so replays feel natural.
You can draft with any chatbot, but rewrite it in your own words. YouTube's monetization policy specifically lists "templated storylines" and "AI-generated content made with generic or unoriginal templates" as ineligible (YouTube channel monetization policies). Your angle, your examples and your opinions are what make it original.
Step 2: Generate the voice-over
Paste the script into the YouTube Shorts voice generator. A few settings matter more than the voice you pick:
- Speed 1.05–1.15× on MiniMax voices gives the quicker Shorts pace without sounding rushed.
- Expressive tags such as
[laughs],[excited]or[whispers]work on ElevenLabs Eleven v3, which is useful for storytime formats. - Use the same voice for a whole series, so returning viewers recognise the channel by ear.
Voice is the cheapest layer. Kenerate prices text-to-speech by characters: MiniMax Speech 2.8 HD is 30 credits per 1,000 characters and Eleven v3 is 60 credits per 1,000 characters. A 60-second script is usually 850–950 characters, so expect roughly 26–29 credits on Speech 2.8 HD or 51–57 credits on Eleven v3. Download the MP3 and note its exact length; you will cut the visuals to it.
Step 3: Plan one shot per sentence
Split the script into lines and write a short visual for each. For a 30-second Short, five or six shots of 5 seconds each is a comfortable rhythm. A simple shot list looks like this:
| Line | Seconds | Shot description |
|---|---|---|
| "This setting doubles your Wi-Fi speed." | 0–5 | Close-up of a router's blinking lights, slow push-in |
| "Most routers ship with it switched off." | 5–10 | Hand holding a phone with a settings screen, shallow focus |
| "Open your router app and find channel width." | 10–15 | Over-the-shoulder shot of a laptop, cursor moving |
| "Switch it from 20 to 80 megahertz." | 15–20 | Macro of a toggle sliding, soft light |
| "Now run a speed test." | 20–25 | Phone screen with a speed gauge rising |
| "Same internet, twice the speed. Follow for part two." | 25–30 | Person relaxing on a couch, phone in hand, warm light |
Writing the shot list first stops you paying for clips you will not use.
Step 4: Make the keyframes (optional but recommended)
Text-to-video works, but image-to-video gives you more control. Generate one still image per shot in 9:16 first, pick the best ones, then animate them. That way you only pay video credits for frames you already like, and a recurring character or product stays consistent across shots.
On Kenerate, Seedream 5.0 Pro costs 13.5 credits per image at 1K, and Nano Banana 2.1 (better at keeping the same character or product across shots from reference photos) costs 21 credits at 1K. Six keyframes on Seedream 5.0 Pro come to 81 credits.
Step 5: Animate the B-roll in 9:16
Open the image-to-video tool, upload a keyframe, set 9:16, and describe the motion in one sentence: subject, action, camera move. Kenerate offers several video models; two are good fits for Shorts:
| Model | Lengths | 9:16 | Credits per second (480p / 720p / 1080p) | Best for |
|---|---|---|---|---|
| Wan 3.0 | 2–30 s | Yes | 15 / 30 / 60 | Value B-roll, drafts, longer single shots |
| Seedance 2.0 | 4–15 s | Yes | 36 / 72 / 180 | Image-to-video ads, dialogue and lip-sync, anime |
Prices are from Kenerate's video model configuration as of 20 October 2026. Two settings save money:
- Turn audio off on the clip if a voice-over will cover it. Both models have an audio toggle, and generated ambience under a narrator just adds noise.
- Draft at 480p, then regenerate only the keepers at 720p or 1080p. Six 5-second Wan 3.0 clips cost 450 credits at 480p and 900 at 720p.
For prompt structure, the Wan 3.0 prompt guide covers subject, motion, camera and sound in detail. Video generation on Kenerate needs a paid plan or credit pack.
Step 6: Assemble the edit
Drop the voice-over and clips into any editor (CapCut, Premiere, DaVinci Resolve or a phone editor). Cut each clip to the length of its line, so the picture changes on the sentence. Keep the first cut inside the first two seconds; a static opening frame is where most viewers swipe away.
Step 7: Burn in captions
Many people watch Shorts on mute, so on-screen text is not optional. Export your edit, then run it through the auto subtitle generator. Kenerate's Auto Subtitles tool transcribes the speech and burns the captions in, with ten styles (including "Rizz", heavy white caps with the spoken word in yellow, and "Glide Pop") and top, centre or bottom placement. It handles clips up to 4 minutes, which covers every Short.
It costs 20 credits for a clip up to 1 minute at 720p, and double that at 1080p. Place captions in the centre or slightly above, because YouTube's own buttons and title cover the bottom and right edges of a Short.
Step 8: Upload and disclose AI use correctly
In YouTube Studio, upload the file and fill in the details. Under Attributes, YouTube asks about "AI use", and you answer Yes or No (YouTube Help). The next section explains which answer is right.
Do you have to label AI-generated YouTube Shorts?
Only when the content looks realistic. YouTube asks creators to disclose content that is AI-generated or meaningfully altered and appears realistic, including when it:
- makes a real person appear to say or do something they did not,
- alters footage of a real event or place,
- generates a realistic-looking scene that did not happen, or
- uses AI-generated music as the main focus of the video.
YouTube also lists uses that do not need disclosure: clearly unrealistic or animated content, beauty filters and colour fixes, AI help with scripts, titles, thumbnails and ideas, captions, upscaling and audio repair, and cloning your own voice for voice-overs (YouTube Help).
In practice, for a faceless Short:
| What's in your Short | Disclose? |
|---|---|
| AI-written script, read by a synthetic narrator over animated or stylised clips | Usually no |
| Photorealistic AI B-roll of street scenes, people or "real" places | Yes |
| AI clip showing a real public figure or a real news event | Yes, and check Community Guidelines first |
| AI-generated song as the main content | Yes |
| Captions and upscaling only | No |
YouTube says disclosing on its own does not limit a video's reach or its eligibility to earn money. Not disclosing is the risk: creators who consistently skip it may get labels applied for them, content removed or a suspension from the YouTube Partner Program. YouTube may also add labels automatically when it detects C2PA metadata or AI generation. When in doubt, tick Yes.
One firm line for this workflow: never generate a real, identifiable person without their consent. A recognisable face saying something they never said is exactly what the disclosure rule exists for, and it can break Community Guidelines on top of that.
Can AI-made Shorts be monetized?
Yes, if they are original. In July 2025 YouTube renamed its "repetitious content" policy to "inauthentic content" and clarified what it covers: content that "looks like it's made with a template", image slideshows or scrolling text with "minimal or no narrative, commentary, or educational value", and AI content made with generic templates (YouTube channel monetization policies). Press coverage at the time stressed that this was not a ban on AI video (Plagiarism Today).
What keeps an AI-assisted channel on the right side of that policy:
- Your own script and angle. A unique take, your research, your opinion.
- Original visuals. Generated clips made for this Short, not stock footage everyone else is using.
- Variety across uploads. A recurring intro is fine; the same storyline with swapped nouns is not.
- No reused clips downloaded from other channels or platforms, even with permission.
If you are still choosing a niche, these faceless AI YouTube channel ideas pair each format with a workflow.
How much does an AI-made Short cost on Kenerate?
Here is a worked example for one 30-second faceless Short with a 60-second budget for the voice (so you can re-record), using prices as of 20 October 2026:
| Layer | Tool and setting | Credits |
|---|---|---|
| Voice-over | MiniMax Speech 2.8 HD, ~900 characters | ~27 |
| Keyframes | 6 × Seedream 5.0 Pro at 1K | 81 |
| B-roll drafts | 6 × 5 s Wan 3.0 at 480p | 450 |
| Captions | Auto Subtitles, ≤1 min at 720p | 20 |
| Total (480p draft) | ~578 |
Rendering the same six clips at 720p instead raises the video line to 900 credits and the total to about 1,028. For scale, the smallest Kenerate credit pack is $15 for 1,300 credits, and new accounts start with 25 free credits, enough to test the voice and image steps before buying anything.
The biggest lever is the number of video seconds. If a shot works as a still with a slow zoom in your editor, skip the video generation for it.
What mistakes make AI Shorts flop?
- A slow first second. Lead with the payoff, and cut to a new shot within two seconds.
- Visuals that don't match the words. One shot per sentence fixes this.
- Generated audio fighting the narrator. Turn clip audio off under a voice-over.
- Captions hidden under the UI. Keep them above the bottom fifth of the frame.
- Horizontal clips. A 16:9 export is not a Short, whatever its length.
- Copy-paste channels. Twenty near-identical uploads are a monetization risk, not a growth strategy.
The short version
Script it yourself, voice it, generate one 9:16 clip per sentence, burn in captions, and answer YouTube's "AI use" question honestly. Keep the voice consistent, keep each Short's idea distinct, and draft cheap before you render high-resolution finals. Start with the YouTube Shorts voice generator: the voice track sets the timing for everything else.
