AI video has received a lot of attention over the last couple of months. While 2023 saw some early iterations of text-to-video, most of the results lacked impact and highlighted a technology that still required more development. This changed when OpenAI unveiled Sora in February this year. The result was a showcase of the potential of a hugely powerful evolution in the video space.
Google also unveiled Veo at its I/O event last month, showing similar results through its own models. Unfortunately, neither Sora nor Veo are available yet to evaluate if the results shown previously are accurate to what is possible.
The latest companies to reveal an advanced AI video platform is Kling AI, which has released its platform to users in China only, and Luma Labs which launched last week and is publicly available to everyone now.
There is also LTX Studios which came out a couple of months ago and has positioned itself as a storyboarding platform that can create serveral short videos. LTX Studio can produce trailers and short films through this technique and even writes voiceover dialogue and music.
Meanwhile, RunwayML and Pika had been the leaders in this space before OpenAI reset everyone’s expectations. Both have continually improved their products over time. RunwayML just announced the third generation of their video model, which will be released in the coming months.
All these platforms are trying to become the market leader in this space, but some are still relatively theoretical as public testing has yet to begin at scale. Similar to the side-by-side comparison of text-to-image platforms previously, let’s do the same with the text-to-video platforms that are currently available, ahead of those we will hopefully get access to in the coming months.
“A ship sailing through a stormy night, lightening.”
“A rabbit wearing a top hat, sitting on a train, drinking coffee”
“a large waterfall in the middle of a serene forest”
LTX Studio
Given the advanced storytelling focus of LTX Studio, it was unfair to do a direct side by side comparision. However, below are two examples of what the platform is capable of producing with the same prompts: