How to Make AI Videos in 2026: Gemini Omni, Flow, Grok Imagine and Midjourney Compared

Turn one photo and one idea into a short video clip in each tool, edit it by talking to it, and know the cost, the watermark and the rules about real people before you post.

Checked October 17, 2026. Researched and drafted with help from Grok Bot.

Quick answer

• AI video is short clips, not movies. Expect a few seconds to about 20 seconds per clip. Longer videos are made by joining clips.
• The easiest start is a photo. "Image to video" (animate a picture you made or own) gives you far more control than text alone.
• Best picks: Gemini Omni for talking to your video and editing it in plain words. Google Flow for building a short film from many clips. Grok Imagine for fast, fun clips right next to your images. Midjourney for art-quality motion.
• You'll need a paid plan. Gemini Omni needs a Google AI plan. Grok Imagine video starts at the Premium plan. Midjourney starts at $10 a month.
• ChatGPT doesn't make video anymore. OpenAI's Sora app was discontinued.
• Real people: only use your own face or someone who said yes. Most apps forbid deceptive videos of real people.

Comparison table

Tool	Where	Start from	Clip length	Edit by chatting	Watermark	Plan
Gemini Omni	Gemini app → Create video	Text, up to 5 images, 1 video	Short clips (see Flow for Omni Flash 4–10 s)	✓ multi-turn edits (objects, camera, scene)	Google AI marking (SynthID)	Google AI plan; 18+
Google Flow	flow.google	Text, images, your clips	Omni Flash: 4–10 s per clip	✓ per clip; build scenes	Google AI marking	Credits: AI Pro 1,000/month
Grok Imagine	Grok app or grok.com	Text or your images	Up to 15 s (SpaceXSI); no separate Premium limit published	✓ in the same thread	Visible Grok watermark (can't be removed)	Premium ($30/mo) or higher
Midjourney	midjourney.com	Your Midjourney image or an image URL	5 s, extend ~4 s up to 4 times (21 s max)	Extend and re-animate	Public gallery unless Stealth (Pro, Mega)	From $10/month
Meta AI	Meta AI app	Your Meta AI images	Very short	Limited	Hidden Content Seal	Free; video details not verified
ChatGPT	✗	✗	✗	✗	✗	Sora discontinued

Which one should I use? The verdict

Quick questions

Can I make AI videos for free? Mostly no. Video uses a lot of computing power, so Gemini, Grok and Midjourney put it on paid plans. Meta AI has some free animation features. Meta AI's free video limits are not verified.

How long can an AI video be? Clips are short: Midjourney is 5 seconds, extendable to 21. Flow's Omni Flash makes 4–10 second clips. SpaceXSI says Grok Imagine makes videos up to 15 seconds, at up to 720p. SpaceXSI doesn't list a separate limit for Premium (not verified). Join clips for longer videos.

Can I remove the watermark? Not in Grok, and you shouldn't try in any app. Removing it breaks the rules of most apps. Google's video is also marked with an invisible SynthID watermark.

Can I put myself in the video? In Gemini Omni you can use your own avatar with @username (after setup). Only use other people's faces with their permission.

Does ChatGPT make videos? Not anymore. The Sora app was discontinued.

Before you start

• Make or pick your starting image. A clear image with one main subject animates best. Make one with our AI Images guide, or use your own photo.
• Plan in seconds. Write what happens in each 4–5 seconds.
• Check the rules. No real people without permission, no logos you don't own, no "news-style" fakes.
• Know your budget. Every clip spends credits or usage. Test ideas as stills first.

The video prompt formula

Video prompts need two things image prompts don't: motion and camera.

[Subject] [doing what], [where], [lighting/mood].
Camera: [static / slow push in / pan left / handheld / drone].
Motion: [slow / gentle / fast], [what moves and what stays still].
Style: [realistic / animated / film look].
Length: [5] seconds. No text on screen.

Example:

A golden retriever runs across a snowy backyard toward the camera, early morning, soft pink light.
Camera: low angle, slow push in.
Motion: dog runs fast, snow kicks up, trees stay still.
Style: realistic, shallow depth of field.
Length: 5 seconds. No text on screen.

Step by step: Gemini Omni

1. Open the Gemini app or gemini.google.com (Google AI plan, 18+).
2. Open Sidebar → Create video. Pick a template or start blank.
3. Add your inputs: up to 5 images and 1 video.
4. Paste your prompt and send.
5. Edit by chatting: Make the camera slower. Change the sky to sunset. Remove the car in the background. Each turn changes the clip.
6. Put yourself in it: set up your avatar, then use @username in the prompt.
7. Share on YouTube or Download.

Where it's limited: uploading a video isn't available in the EEA, Switzerland, the UK and some US states.

Step by step: Google Flow

Flow is Google's filmmaking tool, built for many shots.

1. Go to flow.google and sign in.
2. Start a new project. Add your images (characters, places).
3. Make a clip with Omni Flash (4–10 seconds). Choose resolution: 360p costs half the credits of 720p, so draft at 360p.
4. Add the next shot. Reuse the same character image to keep them consistent.
5. Arrange the clips into a scene and export.

Credits: AI Pro includes 1,000 credits a month, AI Plus 200, and Ultra 10,000 or 25,000 depending on the plan. Without a subscription you get 50 credits a day.

Step by step: Grok Imagine

1. Open Grok (app or grok.com) and open Imagine.
2. Make an image first, or upload your own.
3. Choose to make a video from it and add a motion prompt.
4. Keep going in the same thread: Again, but slower. Same scene at night.
5. Download and share.

Plans: video starts at Premium ($30/month, US, checked Oct 2026). SpaceXSI says Imagine makes videos up to 15 seconds, at up to 720p. SpaceXSI doesn't list a separate limit for Premium (not verified). When you hit your usage cap, 720p video falls back to 480p.

Watermark: a visible Grok watermark you can't remove.

Exact button names in Imagine vary by app version; not verified on an official help page.

Step by step: Midjourney

1. Make an image on midjourney.com.
2. Under the image, choose Animate:
   • Auto (Midjourney writes the motion) or Manual (you write it).
   • High motion or Low motion.
   • Loop to make it loop.
3. Or start from your own picture: add an image URL and --video to your prompt.
4. You get a batch of 4 clips by default (change with --bs).
5. Extend adds about 4 seconds, up to 4 times (21 seconds total).

Plans: video works in Fast mode on all plans. Relax video is on Pro and Mega only. HD video is on Standard, Pro and Mega, in Fast mode.

The others

Try it: a 15-second product ad in 3 shots

1. Write the shot list (any chatbot):

Write a 3-shot, 15-second video ad for [product]. Each shot 5 seconds.
For each: what we see, camera, motion, and a video prompt using this format: [paste formula].
No text on screen; I'll add captions later.

2. Make a hero image of the product (AI Images guide, or your own photo).
3. Make each shot in your tool from the hero image.
4. Join the clips in a free editor (your phone's video app works), add captions and music you have rights to.
5. Label it as AI-generated where the platform asks.

Common mistakes and fixes

Prompt pack

Image to video:

Animate this image. Keep the subject and background exactly as they are.
Motion: [gentle wind in hair, slow blink]. Camera: [slow push in]. 5 seconds. No text.

Fix by chatting:

Keep everything the same, but [make the motion slower / change the time to sunset / remove the person on the left].

Consistent character:

Use this character image for every shot. Same face, clothes and colors. Shot [2]: [what happens].

Loop for social:

Make a seamless 5-second loop: [subject] [gentle repeating motion]. First and last frame identical.

Next level

Want to build a tool instead of a clip? A6: Build a Mini-App Without Code shows Grok Build, Claude, Gemini Canvas and ChatGPT Sites.