First came the AI chatbot, then the AI image generator. Currently, high-tech companies are rushing to release AI video generators.
Over the past year, almost all major tech companies have removed some kind of AI video model. It will soon become a crowded market between Google's VEO 3, Midjourney and Sora. AI videos are a major leap in corporate AI creative offerings, and it is worth noting as generative AI becomes a larger part of the content created and displayed online. This is especially true as technological advancements come when legality, ethics and other concerns surround AI's creative products.
Don't miss out on unbiased technical content and lab-based reviews. Add CNET as your preferred Google source.
Many of these companies also have chatbots and image generators, but I was surprised at how unique differences there are in video generators. Some of my favorite image generator features pop up in the video model, but some are noticeably absent, for example. With so many options, we've put together all the key models that you can consider, including pricing, privacy, and practical testing of each.
This is everything you need to know about AI video generators. This list is updated regularly with the latest information about each generator.
What is an AI video generator?
AI video generators are one of the latest ways high-tech companies use AI generated. These programs use text-to-video image technology that allows you to create short video clips. Enter a short description called a prompt or upload and animate an image, and the software will create a fully created clip with Gen AI. These AI videos tend to be between 5 and 10 seconds, with certain models offering synchronized audio. Because this technology is new, it is possible to make an error called hallucination.
What AI video models are available now?
Examples of AI video generators available today are Sora by Openai, Veo 3 by Google, Adobe Firefly, and V1 by Midjourney. They are all paid programs that produce decent results and allow you to customize your shots with the control panel. Runway, an AI startup that co-created a stable, diffused image generator, is another AI video option with the Freemium plan. Other AI models from Luma, Pika and Ideogram are also available.
Sora joined the ChatGpt family at the end of 2024. This is a rather user-friendly program, but it doesn't have the same conversation UI as Dall-E 3. You cannot “chat” with Sora to create a follow-up revision. Instead, it's more reminiscent of traditional AI creative services. (Disclosure: CNET's parent company Ziff Davis filed a lawsuit against Openai in April, claiming it infringed Ziff Davis's copyright in training and operating AI systems.)
SORA has panels that customize the dimensions, length and stylistic feel of the video. You can enter prompts and upload images for Sora to animate. You can use some editing options to complete the video from there. SORA video is also automatically installed and specified the origin of AI.
SORA is only available to pay for ChatGpt users. For ChatGpt Plus users ($20/month), you can earn 50 priority generation credits per month with up to 5 seconds of video at 720p. Upgrading to Pro Tier ($200 per month) will give you a higher monthly credit, including 500 videos created with priority/fast generation and unlimited videos with relaxed generations. Professional subscribers can also have the option to create high resolution videos at up to 1080p, extend the duration of the video up to 20 seconds, and download the video without watermark.
Openai's privacy policy states that you can train your content unless you opt out. To do that in SORA go to Settings > General and turn it off Improve the model for everyone. You can also exclude videos from the public express feed in your settings.
Google's VEO 3 AI video generator captivated AI enthusiasts when it was released at the 2025 I/O Developers Conference. In my tests, I found that the VEO 3 is a great step up from my disappointing experience with Veo 2, but I can hear the real difference. Veo3 can generate audio in video clips and automatically sync.
VEO 3 video is automatically generated in a horizontal 16:9 format with a resolution of 720p, 8 seconds long. This is a very capable program and while the video is detailed and fun, there are some notable aspects that are not that exciting. VEO 3 is only available on the expensive Gemini Ultra Plan and does not have editing tools. This is a shame. Future updates will make VEO 3 more widely accessible and more sophisticated.
The video has an embedded Google SynthID watermark (not visible in the clip). Google Cloud's general AI policy states that it does not train customer data without permission.
You can also run tests on VEO on YouTube. Thanks to a new feature with VEO, creators can now create AI-generated backgrounds for shorts videos.
Midjourney is one of the most popular AI image generators and has just released its first AI video model called the V1. You can use Midjourney to create video clips between 5 and 21 seconds at 720p resolution. You can use Midjourney using Discord or its website. Currently, video generation is paywalled for users, but it's one of the cheap options for $10 a month.
Midjourney's Privacy Policy states that the personal information and information contained in the prompts can be used to improve our services. When created in stealth mode, the AI images become private. Otherwise, they will be shared in the public gallery.
I was quite impressed with the value for Midi Joanie's price, but you need to be patient and specific to your prompts.
Firefly's standalone AI video generator is currently available on computers and via the Firefly mobile app. If you're familiar with Firefly's AI imaging tools, you'll be familiar with setting up your video generator. The left panel gives you the advantage of being able to customize the clips and selecting the type of motion you want (such as zoom in and out, moving left and right). You can also select the camera angle you want, such as if you want to mimic drone footage in an aerial view.
Some creative cloud plans include Firefly access, such as paying for all of the one program or Adobe apps. You can check and compare options here. If you don't have an existing Adobe plan, you can try out the Firefly Standard Plan ($10 per month) and create up to 20 videos per month. If you need more generation credits, your Pro plan ($30 per month) earns up to 70 videos per month. Both Firefly plans come with unlimited AI image generation. Firefly video is 5 seconds long at 1080p without audio.
Adobe says that videos created with Firefly are commercially safe and that its AI policy does not train with content. Firefly videos do not have any visible watermarks, but their content's credentials are automatically attached to the work. Firefly Models are trained on license domain content and public domain content.
AI enthusiasts may recognize the runway as a startup that co-created the stable spread of popular AI image generators. They may also recognize the runway from a groundbreaking deal they made at major film studios last fall. Lionsgate has agreed to open a catalogue that will be used to create custom AI models for the studio, including thousands of hours of films like the Hunger Games and John Wick, and TV shows like Mad Men.
During a quick test of the service, I was impressed with the quick build tools and the general ease of finding my way. I also previously used the service as part of Canva's Magic Media app. This is useful for Canva Lover. The web app allows you to use Runway for free with 125 credits per month. It's a fairly low limit as it uses around 20 credits for each generation. The upgrade ($15 a month or $144 a year) gives you 625 credits per month, access to new models, and the ability to upscale and download videos towards 4K without watermarking.
Runway's Terms of Use states that you can train AI with prompts and results videos, but do not retain ownership of them. Its privacy policy also states that the runway may disclose your information to affiliate, business and marketing partners. The videos you make are automatically private.
What are your other AI video projects?
There are no particular metas on this list. The company is spending its resources to develop AI, but there are no publicly available AI video generators. I teased the version in October 2024. So far, there are things I know so far.
Meta's AI video model – Movie Gen – is currently just a research concept and is not published.
Thanks to the research paper meta, the movie gen video is 1080p HD, potentially up to 16 seconds long at 16 frames per second. The most notable thing about the film gen is the possibility of synchronized audio. Meta said that film gen can be used to create up to 45 seconds of musical music, ambient noise and instrumental music. It's always possible that this feature won't reach the final cut, but it gives the meta an edge.
You'll probably see that, like Google and YouTube, the first AI-powered features pop up on social platforms, Instagram and Facebook. (There are already many AI features. It's eating space in the feed.) Meta's AI model, chatbot and image generator, is trained with public Facebook and Instagram content, as well as license data.
For more information, check out our guide to writing the best AI image prompts and the best AI chatbots.
