Pick one tool. Write 60 to 80 words. Generate one video. It runs in a browser, there is nothing to install, and from finished script to downloadable file is usually ten to twenty minutes.
The reason you haven't started isn't that it's hard. It's that you opened seventeen comparison articles instead of one tool. Decision fatigue is the actual barrier, and the only fix is to make the decision smaller.
One thing that isn't optional: label the video as AI before you post it. Two taps, and both platforms now issue strikes immediately if you don't.
You typed something into a search bar. Forty-five minutes later you had eleven tabs open, three pricing pages, a Reddit thread from 2023, and nothing made.
I know because I did exactly that. My first evening with AI video, I compared tools for two hours and produced nothing. The second evening I opened one tool and had a video in twenty minutes. The difference wasn't skill. It was that I stopped researching and started making.
So this is the short path. One tool, one script, one video, tonight.
What you actually need
A laptop and reasonable Wi-Fi. That's it.
Every tool worth using at this stage runs in your browser. Nothing to install, no editing suite, no graphics card, no technical background. You can do this sat at the kitchen table at nine at night with a cup of tea, which is where most of my content has been made.
What you don't need, despite what the tutorials imply: a camera, a microphone, a ring light, a studio, a subscription, or a finished brand.
The tools to open first
I'll tell you what I actually use rather than list everything on the market, because the list is the problem.
Open one thing: ElevenLabs
You can do the whole job in here. Images, voice, lip sync. That matters more than any feature comparison, because your problem tonight is that there are too many decisions, and one tool is one decision.
Free tier is 10,000 credits a month, plenty to test with. Paid starts around $6, and the Creator plan at about $22 gives you 100,000 characters.
Design her a voice, don't borrow one
This is the bit most people skip and it takes about two minutes. Rather than picking a stock voice off the shelf, use Voice Design: you write a description of how she sounds, her age, her accent, her tone, her pace, and it generates three versions for you to choose between.
Do that instead of grabbing a library voice, because a stock voice is one that a thousand other accounts are also using. A designed voice belongs to your character. It's the audio half of the same problem as her face, and it costs you nothing extra to get right on night one.
One thing this isn't: cloning. Cloning copies a real person's voice and brings consent paperwork with it. Designing invents a voice that never existed, so there's nobody to get permission from.
For captions, use the app you're posting to
Instagram Edits and TikTok's own editor both do captions and trimming, and they do it inside the app you're already publishing from. That's one fewer thing to download, one fewer thing to learn, and honestly no worse than a dedicated editor at this stage.
What to ignore for now
Everything else. There are dozens of good tools and I use several of them, but not one of them belongs in your first session.
Adding a second tool tonight doesn't give you more options. It gives you another decision, and the decisions are what stopped you last time. Choosing the simpler thing on your first night isn't a compromise, it's the correct call. You can go deeper the moment the basic loop works, and by then you'll know what you actually need instead of guessing.
Skip the blank canvas entirely
If an empty interface is what's stopping you, don't start there. The Avatar Quiz asks you twelve questions about your character, her face, her colouring, her build, her world, then hands you a finished prompt pack: her reference images plus her first podcast and yap stills, ready to paste into whichever tool you've picked. Free, no experience needed, about ten minutes.
Your first video, start to finish
Step 1: Write 60 to 80 words
Word counts make this concrete, so here they are. Thirty seconds is 60 to 80 words. Forty-five seconds is 90 to 110. A full minute is 120 to 150.
Start at 30 to 45 seconds. Short enough to write in ten minutes, long enough to say something real.
Structure: one sentence to introduce, two or three to deliver the actual point, one to close. One idea per video. Three points in sixty seconds gives you something rushed that nobody remembers.
It doesn't need to be good. It needs to sound like a person talking.
Step 2: Generate the voice
Paste your script into ElevenLabs, select the voice you designed for her, generate, and listen to it.
You'll want to change something. Everyone does. Usually it's a sentence that reads fine but sounds stiff when spoken.
Step 3: Generate the video
Feed in your character image and your script, and let it generate the speech and sync the mouth in one pass. Short clips take a few minutes.
Preview before you commit to a long render. Catching a pacing problem on a thirty second preview is much less annoying than finding it after a full export.
Step 4: Caption it
MP4 is the standard export and every platform accepts it. Vertical, 9:16, for Reels and TikTok. Add captions in Instagram Edits or TikTok's editor when you upload, because most people watch with the sound off.
Step 5: Label it as AI, then post
This is the step nobody tells beginners about, and skipping it is the one thing tonight that can actually cost you something.
If your video shows a photorealistic AI person, which yours does, you have to label it. On TikTok there's an AI-generated content toggle in the posting flow. Switch it on and a small badge appears under your username. On Instagram, turn on the AI creator label in your profile settings once and it carries across your account, with a per-post option as well.
Two taps, and then you never think about it again.
Worth knowing why it matters, because "I didn't know" isn't a defence here. Both platforms detect AI content automatically through the credentials baked into the file, so not labelling doesn't mean not being caught. TikTok now issues strikes straight away rather than warning you first: your video comes down and you get a strike, a second one costs you seven days of posting, a third costs you thirty.
Thirty days off posting would end a new account before it started.
This only applies to the photorealistic person. Using AI to help write your caption or your script isn't what these rules are about, and doesn't need labelling.
Then post it.
Writing a script your avatar sounds good reading
This is where the quality actually lives, and it takes no technical skill at all.
- Write how you speak. Contractions. Short sentences. Everyday words. If you wouldn't say it out loud to a friend, cut it.
- Read it aloud before you generate. Every stiff phrase becomes obvious the second you hear your own voice say it. This one habit fixes most bad output.
- Use punctuation for pacing. Commas make short pauses. Full stops make real breaks. Ellipses add hesitation but go unpredictable if you overuse them.
- Spell out awkward words phonetically. If a name or term keeps getting mispronounced, write it how it sounds. Most tools follow the spelling closely.
- Capitals for emphasis, sparingly. One or two words per video at most, or the delivery goes erratic.
What to expect, honestly
Fifteen to thirty minutes from zero to posted once your script is written, a bit longer on a genuine first attempt. Cost on the free tier is nothing. If you decide to pay, one tool is about $22 a month, and you only need that once you know the workflow suits you.
The first video will be fine. Not brilliant. Fine.
The question people ask straight after publishing is always the same: how do I make this look like the same person every time? That's a real problem with a real answer, and it's worth solving before you make forty more. I've written the full method for locking your character down, and it's the piece I'd read next.
And if you'd rather watch someone do it than read about it, the walkthrough lessons inside Human AI take you through the ElevenLabs steps on screen, click by click. They're free, they open the moment you join, and they exist because reading a workflow and doing a workflow are not the same thing.
The short version
- Browser only. No camera, no studio, no technical background.
- ElevenLabs does images, voice and lip sync. One tool, free tier. Captions in Instagram Edits or TikTok.
- 60 to 80 words. One idea. Read it aloud before you generate.
- Label it as AI before you post. TikTok's toggle, Instagram's account label. Two taps, and strikes are immediate if you skip it.
- Ten to twenty minutes from script to file. Fifteen to thirty from zero to posted.
- The first one is meant to be fine, not brilliant.
- Lock your character before you make the next forty.
- Want it shown rather than described? The walkthrough lessons in Human AI are free.
Choose one tool, write one script, make one video today. Not after you've compared another six platforms. The video doesn't need to be brilliant to be worth making. It needs to exist.
And if you get stuck at any point tonight, the walkthroughs are free and they're right there.