Synthesia launches LLM-powered assistant to turn any text file or link into AI video

Synthesia launches LLM-powered assistant to turn any text file or link into AI video


Today, London-based Synthesia, the startup that enables enterprises to create professional AI videos, announced the launch of its AI video assistant, a tool that can turn text-based sources into full-fledged synthetic videos in a matter of minutes.

Available starting today for paying customers, the AI video assistant builds on Synthesia’s existing offerings and can work with any document or web link, making it easier for enterprise teams to create videos for internal and external use cases.

It can easily push enterprises’ content delivery efforts ahead but also draw some eyebrows, given the growing concerns over AI-generated videos and deepfakes — and how they can be used to portray actors and politicians in unexpected ways to mislead internet users.

What can Synthesia AI video assistant do?

Founded in 2017 by a team of AI researchers and entrepreneurs from UCL, Stanford, TUM and Cambridge, Synthesia has been offering an end-to-end platform that allows users to create custom AI voices and realistic-looking AI avatars — or choose from existing ones — and use them to create studio-quality videos.

The offering has drawn significant adoption in enterprise settings, but the process of creating these videos can require some level of effort, including things like writing the script manually (or with AI) and designing the layout of the video by choosing the avatar, language, and elements like logo.

With the new AI video assistant, Synthesia is taking all these steps away. All a user has to do is provide the source of the material — be it a website, text file, word document, PDF or idea — and choose a template with elements like the objective of the video, number of scenes as well as the language and tone of the speaker featured in it. 

“The AI Video Assistant uses generative AI to synthesize the user’s source material together with the user’s objective, audience, and other characteristics. Large language models take these inputs and generate both a script as well as relevant scene layouts from the user’s chosen template,” Sundar Solai, the product manager who led the development of the assistant at Synthesia, told VentureBeat. 

This output is generated within minutes and can then be converted into a video. If the script is not up to the mark, users can stop, go back and regenerate with relevant tweaks in settings to come up with better outputs.

Goal to improve content delivery experience

According to Insivia, people tend to retain 95% of a message when they receive it via video but only 10% when they read it as text.

However, even today, most organizations rely on long, dense handbooks and PDFs when delivering valuable content to their employees and clients. With the new AI video assistant, Synthesia wants to bridge this gap and allow any individual in any organization (with relevant access) to turn disorganized, text-based information into studio-quality videos.

Solai said they have been testing the offering for several weeks now and have noted significant traction from customers. However, he did not share how many companies were part of the early test or how they put the assistant to use.

“In addition to users being able to create videos faster than before, the feature accelerated Synthesia’s ability to make video creation accessible to an even wider audience, including people who previously had limited or no video editing experience,” he noted.

While this can give a big boost to the content delivery efforts of enterprises, it is important to note that the tool is still limited in terms of how much text it can convert into a video. Whether a text file or link, Synthesia points out that there’s a 4500-word source limit on the assistant at present. 

Significant growth despite deepfake concerns

Even as the regulators and the community general continue to debate the risks of AI video and AI voice tools, Synthesia has been going strong with its focus on enterprise users.

The 300-strong company raised $90 million in June 2023 and has roped in more than 55,000 businesses, including half of Fortune 100, as customers. One of those customers is video communication service Zoom, which has been able to create videos 90% faster for sales and training efforts.

“This means Zoom’s subject matter experts no longer need to record themselves, freeing up 15-20 hours of their time each month that they can spend on other tasks as part of their job. Zoom achieved monthly cost savings of $1,000-$1,500 per employee which was previously spent on creating training videos,” Alexandru Voica, the head of corporate affairs and policy at Synthesia, told VentureBeat.

Voica also confirmed that the company is working on making the platform faster as well as the next generation of its avatar technology. This will enable them to show more expressions with more natural movements, better voices and more customization options. This will take the platform’s AI videos a step closer to realistic-looking videos. Other players competing with Synthesia in this space are Deepbrain AI, Rephrase and HeyGen.

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.


Source link