Video generation is now a core component of many AI chatbots, but that doesn’t mean that every app does the task well. In fact, most AI-generated videos sit firmly in the uncanny valley, thanks to myriad errors and distortions. We’ve reviewed the top AI services since their inception and, here, we put them through a series of video generation tests to determine which are best at handling audio, basic scenes, complex motion, and text. ChatGPT’s Sora and Google Gemini’s Veo 3 models are our current favorites since they offer granular control, produce passable audio, and generally create the most realistic clips. That said, all of the apps listed below are worth checking out, so read on for more of our top picks, followed by everything you need to know about AI video generation.
Deeper Dive: Our Top Tested Picks
(Credit: OpenAI)
(Credit: OpenAI)
Pros & Cons
More relevant responses than competitors
Excellent sourcing of up-to-date information
Impressive deep research and writing skills
Best-in-class image generation and recognition
Genuinely useful file processing
Expensive
Occasionally incorrect responses
Why We Picked It
Sora 2 generates believable AI videos of 10 or 15 seconds in length, complete with audio. ChatGPT Pro users can enjoy Sora 2 Pro, which enhances the generation quality and allows for videos up to 25 seconds long. The Pro-exclusive Storyboard feature allows you to plot out different scenes in a video, which is ideal for anything longer and more complex than a meme clip. Perhaps most importantly, Sora’s app, an AI video version of TikTok, is forward-thinking and great for anyone who wants to entertain themselves with an endless feed of AI-generated clips. Although its ability to generate AI videos of real people is a cause for concern, Sora’s Cameo system at least attempts to regulate the practice and allows you to retain control of your likeness.
Who It’s For
Prosumers: If you want to create more advanced AI videos than the average person, ChatGPT’s Pro subscription gets you some significant benefits, including the Sora 2 Pro model and the ability to generate 25-second-long videos.
TikTok fans: If you’re a big TikTok fan or just can’t find enough vertical videos to scroll through, the Sora app has an endless supply of content.
Specs & Configurations
AI Model
GPT-5 and GPT-5 Thinking
Can Resume Prior Conversations
(Credit: Google)
(Credit: Google)
Pros & Cons
Best-in-class value
Capable of complex reasoning, file processing, and web search
Helpful, robust integrations with other Google apps
Paid version includes 2TB of Google One storage
Standout image editing
Occasionally incorrect responses
Inconvenient deep research sourcing
Unimpressive image generation
Why We Picked It
Veo 3.1 has cutting-edge AI video quality and produces accompanying audio. It’s flexible in that you can use more AI credits to create higher-quality videos or fewer to generate as many as possible. Veo’s Flow filmmaking tool is a standout feature, which you can use to extend its eight-second-long clips into longer and more cohesive videos. Other unique features round out Gemini’s video generation offerings, such as its AI animation tool, Whisk, which allows you to turn a picture into an animated still and a still into a full-blown animated clip. More importantly, you need only Gemini’s $20-per-month plan to access Flow, Veo 3.1, and Whisk, rather than its highest-end tier.
Who It’s For
Budget-conscious users: If you want to use best-in-class video generation tools without spending a princely sum, Gemini’s $20-per-month plan lets you spend extra AI credits at once to generate higher-quality videos.
Filmmakers: For filmmakers interested in AI, Veo’s Flow tool lets you create cohesive, cinematic videos.
Specs & Configurations
AI Model
2.0 Flash and 2.5 Pro
Can Resume Prior Conversations
Learn More
Veo 3.1 Review
(Credit: xAI)
(Credit: xAI)
Best for NSFW Videos
Grok
Pros & Cons
Unique features, such as virtual companions and X searches
Good at complex reasoning, file processing, and web searches
NSFW content, if that’s your thing
Expensive
Lackluster image and video generation
Limited research and sourcing abilities
Occasionally incorrect responses
Why We Picked It
Even though Grok generates audio, its video quality isn’t quite at the level of Sora or Veo. That said, it does something that neither of them can: create six-second-long AI NSFW videos for free. From a privacy perspective, Grok’s ability to turn an image of a person (without their consent) into a video in which they take off their clothes is extremely concerning. However, if you’re not using the technology for nefarious purposes, then it’s easy to appreciate it as a tool that gives you more freedom over what you create. If you’re interested in NSFW AI videos, check out Grok’s datable AI companions, too.
Who It’s For
Adults looking for NSFW content: Adults can effortlessly generate NSFW AI videos in seconds with Grok’s Imagine model.
People who oppose censorship: If you’re against all forms of censorship, Grok is much more permissive about what you can generate than competing platforms.
Specs & Configurations
AI Model
Grok 3 and Grok 4
Can Resume Prior Conversations
(Credit: Adobe)
(Credit: Adobe)
Pros & Cons
Good selection of preset styles and effects
Can match the style or structure of uploaded images
Gives you a choice of aspect ratios
Your content doesn’t train Adobe’s AI
Images can look distorted or not match the request
Video creations top out at five seconds in length
No negative prompt option to exclude objects from images
Requires subscription to remove watermark
Why We Picked It
Adobe Firefly has a plethora of features that the other apps on this list don’t, such as an easy and intuitive way to customize a video’s motion, style, and points of reference. It also allows you to generate videos with various models. Furthermore, the video generation process is incredibly quick, and Firefly’s premium plans start at just $10 per month (significantly less than what competitors charge). You can even generate sound effects for videos, using your voice to indicate the perfect timing. However, all these benefits come at the cost of audio sync issues, generally lower video quality than what you get with Sora or Veo, and a maximum clip length of five seconds.
Who It’s For
Tinkerers: With robust customization options, such as the ability to generate sound effects based on prompts, Firefly is great for creators who want control and flexibility.
The privacy-conscious: Adobe says that it doesn’t train its models on your creations, which is great given how much data other AI services collect.
Learn More
Adobe Firefly Review
(Credit: Perplexity/PCMag)
(Credit: Perplexity/PCMag)
Pros & Cons
Top-notch web search abilities
Slick user interface
Variety of models to choose from
Competent creative writing, file processing, image generation, and voice chat capabilities
Not clear which models it uses for media generation
Unimpressive deep research
Why We Picked It
Perplexity doesn’t have a first-party video generation model, but it does use best-in-class technology, courtesy of Gemini’s Veo 3 (for Perplexity Pro users) and ChatGPT’s Sora 2 Pro (for Perplexity Max users). The experience and results aren’t quite as good as if you were to use those models directly on their parent platforms, but Perplexity’s value proposition is slightly better since it provides a competent AI web browser and a class-leading AI web search tool for a lower price ($16.67 per month, billed annually). It’s a good option if you want to dip your toes into the world of AI video generation without forking over the money for a full-on chatbot.
Who It’s For
Casual users: If you have a passing interest in AI video generation, Perplexity allows you to try out top models at a low cost.
People who don’t want to use a chatbot: Chatbots aren’t for everyone. Perplexity distinguishes itself by focuses on AI web search, instead.
Specs & Configurations
AI Model
Sonar and Third-Party Models
Can Resume Prior Conversations
Learn More
Perplexity Review
(Credit: Meta)
(Credit: Meta)
Best for Adding Music
Vibes
Why We Picked It
Meta’s Vibes is more than just a video generation engine; it also features a TikTok-style social platform, similar to Sora. Vibes focuses on short clips of around five seconds in length, which you can customize through different styles and sliders for stylization, variety, and weirdness. This setup makes tweaking videos easy without sacrificing too much control. Unlike most AI video generators, Vibes allows you to add music to videos, giving its creations a distinctly unique feel. Although Vibes’ videos don’t quite look best-in-class, the service works well enough for the more stylized videos you are likely to want to pair with music. Vibes is free, too, so you don’t have to worry about paying for yet another subscription.
Who It’s For
Creators with a particular style: Hyperrealism isn’t always better. Vibes is among the best services we tested for generating cinematic and colorful videos.
Music fans: If you can’t stand videos without music, Vibes lets you easily add a background track to your AI-generated creation.
Get Our Best Stories!
All the Latest Tech, Tested by Our Experts
Sign up for the Lab Report to receive PCMag’s latest product reviews, buying advice, and insights.
Sign up for the Lab Report to receive PCMag’s latest product reviews, buying advice, and insights.
By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy.
Thanks for signing up!
Your subscription has been confirmed. Keep an eye on your inbox!
Buying Guide: The Best AI Video Generators for 2025
How We Choose the Best AI Video Generators
Evaluation how AI video generators handle our test prompts, which you can learn more about below, is a major part of how we decide which apps make this list. Furthermore, all of the entries here are full-service AI video generation apps that work without additional configuration, meaning you can just download them or open their web pages to get started. You can also generate AI videos locally on your computer with a powerful enough setup (such as with Stable Diffusion), but those solutions are outside the scope of this list.
How Do We Test AI Video Generators?
Although we tailor our testing to each app and its unique features, such as ChatGPT’s Storyboard feature in Sora or Gemini’s Flow feature in Veo, our evaluation encompasses four primary pillars: audio, basic scenes, complex motion, and text. We use the exact same prompt to get the clips in the cards above for the sake of a fair comparison.
To test for basic scene generation, including adherence to a given prompt and overall realism, we use the following prompt: “Show me someone going about their daily life in a trendy apartment with rustic decor.”
To test for complex motion, which routinely trips up even the most advanced AI video generation, we use: “Show me a pro Rubik’s Cube solver solving a cube.”
To test for text, particularly while it’s in motion to some extent, we use: “Show me a teacher in front of a class writing down y = mx+b on a whiteboard while explaining the concept.”
When we test the audio component of AI videos, we aren’t analyzing waveforms or measuring bitrates. Rather, if a particular app supports audio generation, we listen closely to see if the sound matches up with what’s happening on-screen. Oftentimes, AI-generated videos with audio lack certain sound effects or include effects that don’t make sense. If the audio matches what’s in the frame, we look to see if it syncs up perfectly, since effects often fire just a moment too soon or late.
Particularly with AI voices, we listen for any distortion or obvious processing. Although it can be somewhat challenging to describe in words, AI voices often have a noticeably robotic quality. A good point of comparison for AI voices is Sesame, a Technical Excellence 2025 award winner; its AI voices sound conversational and exceptionally realistic.
Does AI Video Generation Work Well?
AI-generated videos can sometimes be quite realistic, but you often have to pay for a service, run multiple generations, and carefully tweak your prompt to prevent them from looking or sounding fake. However, the technology is rapidly improving. Soon, it will likely be difficult to tell the difference between an AI video and a real one, even if you use a free app and an unsophisticated prompt.
Do AI Videos Include Audio?
Sometimes, though, this is increasingly common. For example, whereas ChatGPT’s Sora 1 and Gemini’s Veo 2 AI video generation models didn’t generate audio, Sora 2 and Veo 3 do. If you want an AI video with accompanying audio, check if the app you plan to use supports it before refining your prompts.
Is AI Video Generation Free?
Some AI video generation models are free to use, but the answer isn’t always straightforward, even within the context of a single app. For example, ChatGPT initially limited its original Sora model to paid subscribers but made Sora 2 available for free, albeit on an invitation-only basis. If you don’t want to pay for AI video generation capabilities, expect significantly fewer features and restrictive usage limits.
Is There a Length Limit for AI Videos?
Yes, but this limit varies depending on the app. The AI video generation apps we tested don’t support videos longer than 25 seconds, which is itself a limitation, as it is only possible with an expensive ChatGPT Pro subscription. That said, some apps help you work around these limitations. For example, Gemini’s Flow tool in Veo lets you extend AI-generated clips, while Adobe Firefly helps you generate a video based on the last frame of another video you provide.
How Can You Tell If a Video Is Real?
The short answer is that you can’t always tell. A variety of signs can help you determine if a video is likely authentic, but the tell-tale characteristics of AI videos might become increasingly difficult to spot as the technology improves.
Recommended by Our Editors
How Can You Make Better AI Videos?
Prompt engineering is the practice of designing and optimizing requests to an AI to get the best possible results, and this is a key component of AI video generation. Even changing just a few words in a prompt can have a significant impact. Generating something multiple times, even with the same prompt, can also get you different results. Check out our guide on getting the most out of AI video generation for some advice on making the best videos possible.
Can You Make AI Videos of Yourself or Other Real People?
Yes, but there are some caveats. Not every app allows this, and those that do usually have restrictions. ChatGPT’s Sora, for example, features a Cameo system that enables you to generate AI-generated videos of yourself or others that share their likeness. Many historical figures are fair game, too. That said, OpenAI banned the use of Martin Luther King Jr.’s likeness, as well as that of living celebrities (outside of those who set up Cameos).
Can You Make NSFW AI Videos?
Grok’s Imagine AI video generator generates explicit content if requested, but most mainstream AI video models don’t. This feature is more common in the world of NSFW AI chatbots. Grok’s NSFW AI content isn’t particularly impressive, so I suggest using a dedicated service if you’re looking for NSFW AI content.
Disclosure: Ziff Davis, PCMag’s parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
