From Text to TikTok in Seconds: ChatGPT Joins the AI Video Editing Scene

OpenAI’s generative AI is moving into video. Thanks to new multimodal models and plugins, ChatGPT can now turn text prompts into short clips and even tweak existing video content.

For example, OpenAI’s Sora model – a text-to-video system unveiled in late 2024 – “can generate videos up to a minute long” from user instructions. Meanwhile, third-party tools like ByteDance’s CapCut plugin let ChatGPT users create TikTok-style videos with a single sentence prompt. With one short prompt, CapCut’s system produces a full video (complete with auto-generated script, voiceover, music and edits) in minutes​. In effect, ChatGPT can now write and shoot simple videos, without any traditional editing software.

GPT-4’s new “omni” model (GPT-4o) also brings video into the mix. OpenAI says GPT-4o “accepts any combination of text, audio, image, and video” as input, meaning users can soon feed video clips into ChatGPT just as easily as text.

Currently, this is limited to trusted beta testers, but in practice it implies ChatGPT could analyze or transform uploaded videos. In the meantime, ChatGPT’s existing “code interpreter” plugin has shown surprising video chops: users have asked it to trim and convert GIFs, add zoom effects, and output new MP4 clips on the fly​. In short, both by itself and via connected tools, GPT technology can now assist with video creation (text-to-video) and basic editing (e.g. adding transitions or voiceovers).

These developments have generated mixed reactions. Many AI enthusiasts are “amazed” by the new possibilities. Online forums applaud that “nearly anything is possible” – from marketing videos to social media clips – thanks to AI.

NVIDIA researcher Jim Fan hailed Sora as “a much bigger deal” that could eventually replace traditional graphics pipelines. He even suggested that a future version of Sora might become “a full data-driven engine” for video, reducing the need for hand-made 3D animation.

However, prominent skeptics are wary. Meta’s chief AI scientist Yann LeCun slammed the approach as “wasteful and doomed to failure,” arguing that generating pixels frame-by-frame cannot truly understand physics or the real world.

Other critics note that Sora and similar models still struggle with consistency. In hands-on tests, Sora sometimes “refused to generate” a video for a seemingly simple prompt (due to content filters), and even when it did, the results could defy basic logic – for example, a character might fail to blow out birthday candles as instructed, or a moving bat might vanish mid-animation. Such glitches reflect the model’s current limits in understanding motion and causality.

Social media reaction is similarly tempered. Some creators share impressive demo clips made in minutes using ChatGPT and CapCut together, marveling at the ease of turning ideas into video. Others are more cautious: user reports describe Sora’s outputs as “underwhelming” or note that heavy prompting and post-editing are still needed.

Content creators and marketers are excited by how quickly they can produce rough video drafts, but some film and game professionals question the quality and copyright implications of AI-generated footage. One early user survey found excitement about AI’s potential to “revolutionize” fields like film and education, but also fear that raw AI output lacks human creativity and accuracy.

Looking ahead, experts say GPT-driven video tools will have major impacts on content creation. On the plus side, marketers and educators will soon be able to prototype concepts or churn out social media clips far faster.

OpenAI’s own tests with Sora hinted at “world simulation” uses. At the same time, rapid video generation raises concerns about misinformation or deepfakes, just as earlier AI art and voice tools did. Overall, the rise of GPT-powered video is seen as a milestone: a move towards more “natural” AI that can speak, see, and now even film. As one observer put it, the technology is “not the future – it’s happening right now”.