TikTok Parent Company Launches Jimeng AI Text-to-Video App

Aug 15, 2024

A serene nighttime garden scene lit by a bright full moon. Various flowers in shades of blue, purple, and white bloom around a stone path leading to a pond. Tall trees and lush greenery frame the background, with soft moonlight casting a tranquil glow on the foliage.

The parent company of TikTok, ByteDance, has debuted a new text-to-video AI app, Jimeng AI, in China, paving the way for a future where TikTok features a bevy of AI-generated videos.

There is no direct link yet between Jimeng AI and TikTok, but the AI platform creates videos that are formatted for social media by default, leading to some speculation that TikTok may be in Jiameng’s long-term plans. It is worth noting that technically, Jiameng AI is developed by Faceu Technology, which ByteDance owns.

It’s also not the first text-to-video platform to come out of China. As Reuters reports, There’s also Kuaishou’s Kling AI model, which is available in beta form globally. Chinese AI startup Zhipu AI also recently launched a video-generation model, Ying, and another Chinese startup, Shengshu, launched an AI video app called Vidu.

There’s a lot of competition in China, and Jimeng AI also aims to rival well-known text-to-video platforms in the West, including OpenAI’s Sora, which has a dubious training history, Google’s Lumiere, and Pika Labs’ Pika platform. Like Sora, Lumiere, and Pika, Jimeng builds videos based on simple text prompts. In Jimeng’s case, the prompts are written in Chinese, although the reasonable expectation is that the platform will reach other markets at some point. Speaking of Chinese, the root “meng,” means “dream.” Jimeng also includes a text-to-photo generator.

The dedicated Jimeng website outlines precise user control over the text-to-video output and details the app’s pricing structure. Chinese users on Android and iOS can download the app for free, but using the service requires a subscription. For just under $10, users can make nearly 170 videos or more than 2,000 images per month. There’s also an annual plan for around $92.

As for Jimeng’s output, its text-to-video performance looks okay, showing some of the problems common to AI-generated video content, including strange jittering and jilted movement. It isn’t at quite the same level as Sora. The video below was compiled using official Jimeng AI assets by TechRadar.