ChatGPT can't make videos by itself as of September 2026. It makes still images, and OpenAI's video app, Sora, shut down on April 26, 2026. What ChatGPT can do is plan a video, write the prompts, and hand the job to a video plugin from another company, such as Runway, inside the same chat. Below is what ChatGPT does well for a video, how the plugin route works, where Gemini fits, and which AI tools make clips for free.
What ChatGPT can do for a video
ChatGPT handles the planning and the words around the footage, plus the still images a clip can start from:
- Scripts and shot lists. Ask for the video broken into 8-second shots, because that's the longest single clip Veo 3.1 makes. A 30-second ad becomes four shots of about 8 seconds each, which maps onto four generations.
- Prompts for a video model. Ask for each shot as a prompt with a camera move, a subject, one action, the setting and the lighting. Google's Veo prompt guide lists shot framing, style, lighting, character, location, action and dialogue as the parts that give you control.
- Start frames. ChatGPT makes and edits still images, and its September 8, 2026 update, ChatGPT Images 2.5, added templates, plus sketch-to-image on mobile. A still you like can become the first frame of a clip in a video tool that accepts a start image.
- Captions, titles and voiceover text. Once the clips exist, ChatGPT can write the on-screen captions and a read-aloud script that fits the length.
A request that works well in ChatGPT reads like this: "Write a 4-shot plan for a 30-second video of a bakery opening at dawn. For each shot, give one prompt of 40 to 60 words with the camera move, the action, the setting and the light, plus any spoken line in quotes."
Where ChatGPT falls short for video
ChatGPT has no video model of its own since Sora ended. OpenAI's help center says the Sora app and website shut down on April 26, 2026, and it set September 24, 2026 as the end of the Sora API. Former Sora users can export their content with the Export button on OpenAI's Sora sunset page, and OpenAI says the data will be permanently deleted after any final export window.
If you're hoping for a setting that makes ChatGPT produce videos, there isn't one. ChatGPT Images, including the September 2026 update, makes still pictures. A plugin fills the gap, but the clip's length, price and sound come from the plugin's own model and plan, so compare those rather than your ChatGPT plan.
How to make a video inside ChatGPT with a plugin
A ChatGPT plugin connects a chat to another company's service, and several video makers have one. As of September 2026, ChatGPT's plugin directory lists Runway ("Generate with every AI model"), Higgsfield ("Every image and video model"), invideo ("Create videos of any length") and Viewmax. Runway's listing says it works through your Runway account, keeps results in your Runway workspace, and sends any purchase to Runway's own site instead of taking payment in the chat.
These steps follow OpenAI's plugin page:
- Open the ChatGPT sidebar or ChatGPT's plugin directory, and find the video plugin you want.
- Add the plugin and review the data and actions it asks to use. On Enterprise and Edu workspaces, an admin has to enable plugins first.
- Connect your account with the video company if the plugin asks for it.
- In a chat, type "@" and the plugin's name, or pick it from the Tools menu, then describe the video.
OpenAI says select plugins are available on all ChatGPT plans. Before you add one, look up its price per clip on the company's own pricing page. Runway's Free plan, for example, is a one-time 125 credits.
Can ChatGPT make videos from images?
Not by itself. ChatGPT can describe a photo and write a prompt for animating it, but the animation happens in an image-to-video tool or a plugin. These accept a starting image:
| Tool | How you add the image | Clip length | Needs |
|---|---|---|---|
| Runway plugin in ChatGPT | Upload the photo in the chat and ask Runway to animate it | Set by the Runway model you use | A Runway account and credits |
| Gemini app (Gemini Omni) | Add image, up to 5 photos | 10 seconds | A Google AI plan, age 18 or older |
| Google Flow (Veo 3.1) | Frames: a start frame and an optional end frame | 4, 6 or 8 seconds | Free daily credits or a plan |
| Midjourney | Animate Image, on one of your Midjourney images or an uploaded photo | 5 seconds, extendable to 21 | A Midjourney subscription |
| Toybox Video Creator (Veo 3.1) | Start from a photo, plus an optional end frame | 4, 6 or 8 seconds | A Toybox Pro plan or higher |
A photo that moves well has one clear subject and room around it for the motion. In a crowded shot it's hard to say in a prompt what should move, so start with a simple one. The image to video guide goes deeper on animating a single photo.
Can Gemini make videos?
Yes. Gemini's video feature runs on Gemini Omni, the model Google now uses in place of Veo in its Gemini app. It's for personal accounts with a paid Google AI plan, and only for people 18 and older.
- In the Gemini sidebar, choose Create video (templates are optional).
- Describe the clip you want. Attach up to 5 photos with Add image, or one clip with Add video, if it should build on your own media.
- Click Submit and give it a few minutes.
- Save the result from Share, using Download video.
Google says each Gemini Omni video runs about 10 seconds with sound generated along with it, and follow-up messages in the same chat can swap a character, adjust the lighting or change the background. Veo 3.1 itself, with its cheaper Lite model, is a choice in Google Flow instead; the Veo 3.1 how-to lists Flow's steps and daily free credits.
Can AI make a video for me?
AI can make short clips for you, and a finished video is still something you assemble. As of September 2026, video models generate one shot at a time: Veo 3.1 makes 4 to 8 seconds, Gemini Omni up to 10, and Midjourney 5. Some add sound, and Google documents Veo 3.1 audio as native, with spoken lines written in quotes.
A 30-second or longer video comes from one of three approaches:
- Several clips, edited together. Write a shot list, make one clip per shot, then join them in any video editor with captions and music.
- Extending a clip. The Gemini API adds 7 seconds to a Veo 3.1 clip per extension, up to 20 extensions. Midjourney adds 4 seconds per extension, to a 21-second maximum.
- An all-in-one platform. Runway's paid plans, from $15 a month, include several AI video models, plus tools for music, voiceover and sound effects, in one account.
The step-by-step version of the first approach is in how to make AI videos. For other jobs AI can take off your hands, from pictures to photo edits, see what AI can make for you.
Is there an AI that can create videos for free?
Yes, within limits. These free options exist as of September 2026:
- Google Flow. Flow adds 50 credits to each account every day, and Veo 3.1 Lite takes 10 per clip, enough for five Lite clips daily. Without a plan, video generation can be unavailable at Google's peak time, around 2 PM to 5 PM UTC.
- YouTube Shorts. In September 2025, YouTube announced free AI clips, sound included, for Shorts, starting in five countries: the United States, Canada, the United Kingdom, Australia and New Zealand. To find it, tap the create button and look for the sparkle icon at the top right. Google DeepMind now lists YouTube Shorts as a place to try Gemini Omni, so the model behind the feature may have changed.
- Runway's Free plan. A single grant of 125 credits for trying Runway's tools, with no monthly refill.
- Open-source models on your own computer. Wan 2.2's models carry the Apache 2.0 license. Its smaller 5B model outputs 720p at 24 fps if your graphics card has 24 GB of memory or more, such as an RTX 4090. The software is free; the hardware and the setup are on you.
Toybox isn't a free option: you can sign up free and get 20 credits, and AI images, flyers, videos and songs need paid credits or a plan.
Make a short clip in Video Creator
A ChatGPT shot list can go into Toybox AI's Video Creator one shot at a time. Video Creator is powered by Google Veo 3.1 and needs Pro or higher. A clip currently costs 350 credits, and the Lite version, which uses Veo 3.1 Lite, costs 180 at 720p or 220 at 1080p.
- Copy a single shot from the list into the "Describe your video" box.
- If the shot opens on a particular picture, such as a still you made in ChatGPT, attach it with "Start from a photo".
- Match "Format" to where it will play (Vertical for Reels and Shorts, Wide for YouTube) and set "Length" to the shot's seconds.
- Press "Create video" and wait a few minutes. A clip that fails gives its credits back.
- Save each MP4, then line the shots up in your editor.
A single shot written for it looks like this:
Slow push-in on a small corner bakery at dawn. A baker in a white apron flips the door sign from Closed to Open, then steps back and smiles at the empty street. Warm light spills from the window onto wet cobblestones. Soft blue morning light, handheld documentary style, shallow depth of field.Video Creator clips come with sound. There's no separate setting for it, so add a sound line to the prompt, such as "Ambient noise: birds and a quiet street." To use your own voiceover or music, add it as part of the edit.