How do I generate video with AI? A step‑by‑step guide for 2026

Here is the short version. To generate video with AI, you pick a tool that fits the job, write a prompt that describes one shot (subject, action, setting, camera, light), set the length and aspect ratio, generate a few versions, and stitch the best clips together in an editor. Most models produce clips of 5 to 10 seconds, so a 30-second video is really four to six generations joined up.

This guide walks through that process step by step, with copy-paste prompts and one worked example with real credit numbers. We compiled it from official product pages and published pricing breakdowns, checked as of late October 2026. Nobody at the site ran hands-on tests for this piece. If you want to compare tools first, see our list of the best AI video generators.

What you need before you start

• A clear goal. "A 15-second vertical product teaser for Instagram" is a goal. "A cool video" is not.
• A shot list. Three to six lines, one per clip. Each line becomes one prompt.
• An account on one generator. Free tiers exist, but they are small. See step 2.
• A basic video editor. CapCut, DaVinci Resolve, iMovie or the editor inside your generator. You need it to join clips and add sound.
• Reference images (optional). A product photo or a style frame gives the model something concrete to match.

One change to know about: OpenAI announced on 24 March 2026 that it would discontinue Sora in the consumer app and API. The Decoder reports the web and app closed on 26 April 2026 and the API on 24 September 2026. Older tutorials that start with "open Sora" no longer work. Our Sora review covers what happened.

Step 1: Pick the right type of AI video tool

"AI video" covers three different jobs. Picking the wrong type is the most common reason people give up.

Type	What it does	Examples	Use it for
Text-to-video / image-to-video model	Generates new footage from a prompt or a still image	Gemini Omni, Kling, Runway	Cinematic shots, product scenes, b-roll
Avatar video	A digital presenter reads your script	HeyGen, Synthesia	Training, explainers, sales messages
Template or script-to-video editor	Turns a script or article into edited scenes with stock clips and captions	Pictory and similar tools	Social clips, repurposing blog posts

Example: if your video is a person explaining your refund policy, use an avatar tool (see our HeyGen review). If it is a coffee cup on a rainy windowsill, use a text-to-video model. The rest of this guide focuses on text-to-video, because that is what most people mean by "generate a video with AI".

Step 2: Choose a free or paid route

Free tiers are good for learning, not for volume. Here is what the main options offer, as reported by the sources linked in each row.

Option	Free allowance	Main limits	Cheapest paid step
YouTube Shorts / YouTube Create (Gemini Omni Flash)	Free with a YouTube account	Clips capped at 10 seconds and 1080p	Google AI Pro, $19.99/month with 1,000 Flow credits
Kling AI	66 credits a day, expiring after 24 hours	360p to 540p, watermark, no commercial use	Standard, $10/month for 660 credits
Runway	125 one-time credits (about 25 seconds)	Watermark, credits do not refill	Standard, $15/user/month billed annually
HeyGen (avatars)	3 videos a month	Avatar format only	See vendor pricing

The Gemini app itself is no longer a free route. A promotion gave free users 10 videos, but it ended on 4 August 2026, and Google's plan page now lists video generation as a feature of paid tiers. For a wider list, see our best free AI video generators.

Step 3: Write a structured prompt for one shot

Think of a video prompt as a camera brief, not a story. The model renders one continuous shot, so describe one shot. Use this five-part template:

[Subject] [doing action], in [setting]. [Camera: shot size and movement]. [Lighting and time of day]. [Style or film look]. [Audio, if supported].

Example prompt (copy and paste):

A white ceramic coffee cup steaming on a wooden windowsill, rain running down the glass behind it. Slow push-in from a medium shot to a close-up. Soft grey morning light, warm reflections on the cup. Shot on 35mm film, shallow depth of field. Quiet rain sounds.

Keep it to 40 to 80 words. Name concrete objects and one clear motion. "Beautiful, cinematic, amazing" adds nothing the model can act on. Our guide to AI video prompts has more templates.

Step 4: Set length, aspect ratio, resolution and audio

These settings decide your cost, so set them before you press generate.

• Aspect ratio: 9:16 for Shorts, Reels and TikTok. 16:9 for YouTube and websites. 1:1 for feeds.
• Length: start with 5 seconds. Longer clips cost more and drift more.
• Resolution: draft at the lowest setting, upgrade only the takes you keep.
• Audio: native audio usually costs extra. On Kling's VIDEO 3.0, a 5-second clip costs 30 credits at 720p without audio and 60 credits at 1080p with audio, double the price for the same shot.

Example settings: 9:16, 5 seconds, 720p, audio off. Add sound later in the editor.

Step 5: Generate, review and change one thing at a time

Generate two or three takes of the same prompt. Then check each one against a short list: does the subject stay the same shape, do hands and text look right, does the camera do what you asked?

If a take fails, change one variable and regenerate. Change the camera move, or the lighting, or the action, but not all three. If you change everything at once you cannot tell what fixed it. Opus Clip's Gemini Omni guide gives similar advice: start broad and refine over several turns, asking for targeted changes while keeping the rest of the scene the same.

Example follow-up prompt:

Same shot, same cup and window. Change only the camera: static close-up instead of a push-in.

Budget for this. Kling and Runway both charge for failed or unwanted generations, according to the pricing breakdowns from eesel and Creatify.

Step 6: Edit, add sound, export and label

Drop your best clips into an editor in shot-list order. Trim the first and last half-second of each clip, where motion often looks unnatural. Add music, a voiceover (our guide on adding AI voice to video covers this) and captions. Export at the platform's native ratio.

Then label it: check each platform's disclosure rules for realistic AI content before you post. Google also embeds an invisible SynthID watermark in every Gemini Omni video. Viewers are getting better at spotting AI footage too; our explainer Is this video AI? shows what they look for.

Example: a complete worked example

Goal: a 15-second vertical teaser for a small coffee brand, built on Kling's free tier.

Shot list (three 5-second clips):

Shot 1: Close-up of roasted coffee beans pouring into a grinder, slow motion, warm side light, 35mm film look.

Shot 2: A white ceramic cup steaming on a wooden windowsill, rain on the glass, slow push-in, soft morning light.

Shot 3: A hand lifts the cup out of frame, static medium shot, same windowsill and light as before.

Settings: 9:16, 5 seconds, 720p, no audio.

Credit math: at 30 credits per 5-second clip, one take of each shot costs 90 credits. Add one retry per shot and you need 180 credits. The free tier gives 66 credits a day that do not roll over, so that is two clips a day, or about three days of free generation. Free output is also capped at 360p to 540p with a watermark, per eesel's breakdown.

On the $10 Standard plan (660 credits), the same 180-credit project uses about 27% of a month's credits, and you could make about 22 clips of this type in total.

Finish: join the three clips in an editor, add a 15-second music bed and one caption ("Slow mornings, roasted fresh"), export at 1080x1920 and mark it as AI-generated when you upload.

Common mistakes and how to fix them

This is where AI video breaks most often.

Problem	Likely cause	Fix
Subject morphs or melts mid-clip	Clip too long or too much motion	Cut to 5 seconds, ask for one simple movement
Character looks different in every shot	Each generation starts fresh	Use image-to-video with the same reference image for every shot
Garbled on-screen text	Models still struggle with lettering	Add text in the editor, not in the prompt
The model ignores half the prompt	Prompt describes several scenes	One shot per prompt, split the rest into new prompts
Credits gone in one evening	Drafting at full resolution with audio	Draft at 720p without audio, upgrade only the final takes

Tips to get better results

• Start from an image. Image-to-video gives you control over composition and keeps products and faces consistent. See our image-to-video AI tools roundup.
• Use film language. "Medium shot", "dolly in", "overhead", "golden hour" are terms models were trained on.
• Write in English for predictability. Opus Clip's guide notes English prompts reportedly give the most predictable results.
• Keep a prompt log. Save every prompt with its settings and a note on what worked. It turns guesswork into a repeatable recipe.
• Check rights before commercial use. Kling's free tier, for example, carries no commercial usage rights.

Next steps

Run the worked example once on a free tier to learn the loop: prompt, settings, review, revise, edit. Once you know how many retries a typical shot takes, you can estimate a monthly credit budget and pick a plan. If you need a presenter instead of footage, try an avatar tool. For a ranked comparison of paid options, go back to our AI video generator roundup.

FAQ

How can I use AI to generate a video?

Pick a text-to-video tool such as Gemini Omni, Kling or Runway, write a prompt that describes one shot (subject, action, setting, camera, light), and set the ratio and length. Generate a few takes, keep the best and join several 5 to 10 second clips in a video editor, adding music, voice and captions.

How do I generate video with AI for free?

Use a free tier. As of October 2026, YouTube Shorts and the YouTube Create app offer Google's Gemini Omni Flash at no cost, Kling gives 66 credits a day, and Runway gives 125 one-time credits. Expect watermarks, low resolution or short clips, so free tiers suit drafts and learning rather than client work.

How do I generate video creation with AI?

Treat it as a pipeline rather than one prompt. Write a short shot list, turn each line into a prompt, generate each clip, then assemble them in an editor. For talking-head videos, an avatar tool such as HeyGen or Synthesia reads your script instead. For repurposing articles, a script-to-video editor builds scenes automatically.

How to generate a free video with AI?

Open YouTube Create or a Kling account, choose text-to-video, paste a one-shot prompt, and set 5 seconds at the lowest resolution to save credits. Generate two takes, download the better one and repeat for each shot. Join the clips in a free editor such as CapCut or DaVinci Resolve, then label the result as AI-generated.

How to generate AI videos for free?

Spread the work across free allowances. Kling's daily credits reset every 24 hours, so plan two short clips a day; use YouTube Shorts for quick vertical clips; and save Runway's one-time credits for a shot that needs them. Draft at low resolution without audio, which roughly halves the credit cost per clip.