How live commerce video production was automated using only product URLs, without the need for show host filming

About

As live commerce has established itself as an essential sales channel for beauty and commerce brands, there is a growing demand to quickly produce broadcasting content for every new product launch and promotion. However, traditional production has relied on repetitive tasks—planning and writing scripts from scratch for each broadcast and coordinating show host casting and filming schedules—which limited production speed and content scalability.

To solve this, we introduced a live commerce video auto-generation solution based on Perso Interactive. By entering a product URL and custom prompts, the AI automatically generates a script for broadcasting, creating a draft video in a customized format tailored to the brand's exclusive voice and channel.

As a result, the scriptwriting process shifted from starting on a blank slate to beginning with an AI-generated draft. With the addition of multi-language support and scene-by-scene editing capabilities, a production foundation has been established to expand commerce content in large volumes and multiple languages.

Background

While the share of live commerce continues to grow, brands must constantly produce broadcasting content for every new product launch and promotional event, placing a significant burden on production teams. For every broadcast, they had to plan and write scripts from scratch—from opening to closing—and substantial resources were consumed in casting show hosts, coordinating filming schedules, and handling post-production editing. As the number of products to be covered increased, it became structurally difficult to produce high volumes of broadcasts of consistent quality at a rapid pace.

Moreover, expanding into global markets added the challenge of having to recreate content for each language. This implementation started with a simple question: 'Can we quickly generate a draft of a broadcast video just by inputting product information, without having to shoot?'

Solution

By simply entering a product URL and custom prompts, this solution automatically generates broadcast scripts for live commerce. The core advantage is that draft videos can be obtained quickly without requiring a show host to film in person.

To eliminate the need to restructure the script every time, it is provided in a standard broadcast template format: Opening Mentions → Brand and Product Description → Promotion → Closing. In addition, an pre-recorded brand-exclusive voice can be selected from a TTS library and applied, ensuring a consistent brand tone across all broadcasts.

Videos are delivered in customized formats optimized for channels like live commerce and short-form videos. Supporting up to 6 languages in addition to Korean and English, multi-language broadcasting content can be produced within the same workflow. It also supports scene-by-scene editing and partial exports, allowing modifications—such as swapping out promotional segments—to be processed without re-rendering the entire video. The duration for each section can be set anywhere from 1 to 20 minutes.

Operations follow a hybrid workflow combining AI automated generation and final editing by a human creator. The production team finalizes the content during the editing stage by inserting product images or video materials and performing fine-tuning on the duration. This creates a structure where the AI builds the draft and a human refines it to perfection.

Results

Following the introduction of this solution, the draft stage—previously the most burdensome part of the production process—was replaced with an AI-generated starting draft, significantly streamlining the overall planning and writing workflow. Thanks to the brand-specific voices and structured templates, the broadcasting tone has been standardized, and video content can now be delivered in customized formats optimized for various channels, including live commerce and short-form videos. In addition, multi-language support has paved the way to scale up into global commerce content, and scene-by-scene editing makes minor adjustments like promotional updates easy to manage.

Recommended for

This case is highly beneficial for commerce and beauty brands that frequently produce live commerce and short-form content, as well as marketing and content teams need to make a high volume of broadcast content with limited resources. It will serve as an excellent reference for video production teams seeking to reduce resources spent on casting show hosts and filming while increasing production speed, and for global brands in need of multi-language commerce content targeted at overseas markets.

Frequently Asked Questions

Q. Is the video fully completed right away without any separate filming?

A. The script and the speaking video of the AI Avatar are generated automatically. However, since product images and videos are not automatically inserted into the background, the brand needs to prepare those materials and insert them manually during the editing phase. In other words, it operates as a hybrid format of AI automation + final editing by the creator.

Q. Can the exact video length or duration be precisely matched?

A. The duration of each section can be set from a minimum of 1 minute to a maximum of 20 minutes. However, because precise automated control down to the character count or second has limitations due to the nature of LLMs, the final length is adjusted by the creator during the editing stage.