Video Editor
TikTok
Design tools like Figma support long sessions, while social media content tools cater to quick interactions. Users focus on posting, not editing, so our goal is to streamline that process efficiently.
1. Details make a huge difference due to large scale of user base
We conducted numerous A/B tests on TikTok to identify designs and features that affect posting rates. Surprisingly, small changes often impact outcomes more than new features, as they shape users' initial perceptions of functionality.
A/B Test: Make the main posting action easily accessible, as users are here to post content, not edit.
A/B Test: On the camera or video screen, filled icons are better recognized due to the constantly moving and changing background.
A/B Test: Small change of copy can significantly increase posts from albums, showing how minor wording differences can greatly influence user understanding.
2.If possible, minimize the effort users spend on editing. For example, when I designed the Auto-Caption feature, I spotted where the effort lies first and let AI handle the task.
With AI's help, we can generate subtitles and insert them into the timeline. After demo testing with our engineers, we found that timing accuracy is higher than word recognition accuracy. This led us to think that maybe we don’t need a timeline.
Get insights from user behaviors
I interviewed several creators to observe how they add captions. One creator stood out by first writing down notes for each subtitle line and then copying them into the video editor.
This is what creators do before they using the editor to add captions.
This inspired me to think—maybe a timeline isn’t always essential for video editing. Instead, we can let users edit captions as seamlessly as editing text.
3.Design beyond the interface. Just like the auto-cut feature, we need to understand the mechanics behind a seamless user experience.
This feature was built long before LLM but relies on the same AI principles. If you don't design how the tool works, you won't achieve satisfying results, especially with effect tools.
This feature offers a seamless UX: users simply upload multiple video clips and choose a soundtrack, and our system automatically edits them to match the music’s rhythm, creating a well-paced and engaging video.
4.The tool’s UX should be incredibly simple, yet it must offer enough editing capabilities to avoid feeling too limited.
This feature was built long before LLM but relies on the same AI principles. If you don't design how the tool works, you won't achieve satisfying results, especially with effect tools.
In video editing, overly simple features limit creative potential. So, we upgraded the trimming tools for users to play with their clips and create something fun.
For features that may overlap, such as voiceovers and sound effects, it’s essential to ensure they blend seamlessly with each other.
5.In social media, the Feed always attracts the most traffic. We should make sure the tools available in the camera or video editor are easily discoverable upfront.
This feature was built long before LLM but relies on the same AI principles. If you don't design how the tool works, you won't achieve satisfying results, especially with effect tools.
People are often drawn to what others have. We include various prompts in the Feed to reduce decision-making for those who are hesitant to give it a try.
