The rise of generative AI for video. A new website, QuickVid, combines multiple generative AI systems into one tool to automatically create short-form YouTube, Instagram, TikTok, and Snapchat videos.
In just one word, QuickVid selects a background video from its library, writes a script and keywords, overlays images generated by DALL-E 2, and synthesizes narration and narration from YouTube’s royalty-free music library. Add background music. QuickVid creator Daniel Habib says he’s building the service to help creators meet the “growing” demand from fans.
“By giving creators the tools to create high-quality content quickly and easily, QuickVid helps creators increase content output and reduce the risk of burnout,” Habib told TechCrunch. said in an email interview. “Our goal is to leverage advances in AI to empower our favorite creators to meet the demands of their audience.”
But tools like QuickVid, depending on how you use them, can flood an already crowded channel with spam and duplicate content. You may also face pushback from creators who choose not to use the tool due to cost ($10/month) or principle, but may have to compete with a large number of new AI-generated videos. There is a nature.
chase the video
Built in a few weeks by Habib, a self-taught developer who previously worked at Facebook Live and video infrastructure Meta, QuickVid launched on December 27th. January — However, QuickVid can combine components that make up a typically informative YouTube Short or TikTok video, such as captions and avatars.
Usage is simple. First, the user enters a prompt describing the subject of the video they want to make. QuickVid uses prompts to generate scripts and leverages the generated text feature of GPT-3. From keywords automatically extracted from the script or manually entered, QuickVid selects a background video from our royalty-free stock media library Pexels and uses DALL-E 2 to generate the overlay image. It then outputs the narration via Google Cloud’s text-to-speech API. — Habib says users will soon be able to replicate their voice — before combining all these elements into a video.
Image credit: quick video
See this video made with the prompt “Cats”.
or this:
QuickVid isn’t pushing the boundaries of what’s possible with generative AI. Both Meta and Google are showcasing AI systems that can generate completely original clips given a text prompt. But QuickVid fuses existing AI to take advantage of the repetitive, templated format of B-roll-heavy short-form videos, avoiding the problem of having to generate the footage itself.
“Successful creators have very high quality bars and are not interested in publishing content that is not their opinion,” Habib said. “This is the use case we are looking at.”
If so, QuickVid’s videos are generally mixed in terms of quality. Background videos tend to be a little random or just tangent to the topic. This is not surprising given that QuickVids is currently limited to the Pexels catalog. It shows the limits of technology.
In response to my feedback, Habib said QuickVid is “tested and tinkered with daily.”
copyright issues
According to Habib, QuickVid users retain the rights to commercially use the content they create and have permission to monetize it on platforms such as YouTube. But the copyright situation with respect to AI-generated content is vague…at least for now. The United States Patent and Trademark Office (USPTO) recently moved to revoke copyright protection for AI-generated comics.
When asked about how the USPTO’s decision would affect QuickVid, Habib said it only concerns the “patentability” of AI-generated products and the rights of creators to use and monetize their content. I said I don’t think it matters. As he pointed out, creators don’t often file patents on their videos, usually leaning toward the creator economy, allowing other creators to reuse clips to expand their own reach.
“Creators are interested in having high-quality content in their voice that can help grow their channel,” says Habib.
Another pressing legal issue could affect QuickVid’s DALL-E 2 integration. This may in turn affect the site’s ability to generate image overlays. Microsoft, GitHub, and OpenAI are being sued in a class action lawsuit, alleging code generation system Copilot violated copyright law by allowing sections of licensed code to be regurgitated without providing credit. . (Copilot was co-developed by Microsoft-owned OpenAI and GitHub.) This case has implications for generative art AI like DALL-E 2. Similarly, DALL-E 2 has been found to copy and paste from its trained dataset (i.e. images).
Habib isn’t worried. The generative AI genie claims to be out of the bottle. “If another lawsuit comes up and OpenAI disappears tomorrow, there are some alternatives that could power QuickVid,” he said, referring to his open-source DALL-E 2-like system Stable Diffusion. said. QuickVid has already tested his Stable Diffusion for generating avatar images.
Moderation and Spam
Legal dilemmas aside, QuickVid may soon have moderation issues. OpenAI implements filters and techniques to prevent them, but generative AI has well-known toxicity and factual accuracy issues. GPT-3 spits out misinformation beyond the boundaries of its knowledge base, especially about recent events. ChatGPT, a fine-tuned descendant of GPT-3, has also been shown to use sexist and racist language.
This is especially a concern for those who use QuickVid to create informational videos. In a quick test, I had a partner (especially a much more creative person in this space than I am) type in some offensive prompts and see what QuickVid produces. To QuickVid’s credit, apparently problematic prompts like “Jewish New World Order” and “9/11 conspiracy theory” didn’t produce toxic scripts. But for “critical racial theory indoctrinating students”, QuickVid generated a video that hinted that critical racial theory could be used to brainwash schoolchildren.
look:
Habib said he relies on OpenAI’s filters for most of his moderation work, and manually reviews all videos made by QuickVid to ensure “everything is within the law.” It is the user’s duty to do so.
“In principle, I think people should be able to express themselves and create content they like,” Habib said.
It appears to contain spam content. Habib argues that the algorithms on his platform, not QuickVid, are in the best position to judge video quality, and that anyone who creates low-quality content “just hurts their reputation.” increase. Reputational damage would naturally discourage people from using his QuickVid to create large-scale spam campaigns, he says.
“If people don’t want to see your video, you can’t get it delivered on platforms like YouTube,” he added. “Producing low-quality content will make people view your channel negatively.”
But it’s instructive to look at ad agencies like Fractl, which in 2019 used an AI system called Grover to generate entire sites of marketing materials. In an interview with The Verge, Fractl partner Kristin Tynski said he foresaw generative AI enabling “massive tsunamis of computer-generated content across every imaginable niche.”
In any case, video-sharing platforms like TikTok and YouTube didn’t have to contend with moderating AI-generated content at scale. Deepfakes — synthetic videos that replace an existing person with someone else’s caricature — began being introduced to platforms like YouTube a few years ago with tools that made it easy to create deepfake footage. But unlike today’s most compelling deepfakes, the types of videos QuickVid creates are clearly not AI-generated.
Google Search’s policy on AI-generated text may be a preview of what’s to come in the video domain. Google does not treat synthetic text differently than human-generated text when it pertains to search rankings, but does not treat synthetic text as content “intended to manipulate search rankings and not be useful to users.” We will take action against it. This includes the[doesn’t] Content that not only adds “sufficient value” but is generated through a purely automated process may also apply to QuickVid.
In other words, AI-generated videos may not be outright banned from platforms, but simply a cost of doing business. While this may not allay the concerns of experts who believe that platforms like TikTok are becoming the new hub for misleading videos, as Habib said in an interview, it will “power the generative AI revolution.” You can’t stop it.”
