After recording an hour-long podcast or lecture, picking out ten highlight segments suitable for Reels takes about two or three hours to do by hand. Latte Social uses AI to analyze a long video's content rhythm, semantic peaks, and emotional ups and downs, automatically picking segments likely to go viral, directly outputting them as vertical short clips, and even auto-adding subtitles and reframing.
Features and use scenarios
The core flow is: upload long video → AI marks climax segments → auto-crop to 9:16 vertical → generate subtitles → smart auto-focus (keeping the speaker centered in frame). The smart auto-focus feature is especially useful for two-person conversations, no need to manually adjust framing segment by segment.
It suits podcast hosts, online-course instructors, enterprise marketing teams, and content-outsourcing teams. Taiwan's creators making knowledge content have long been stuck at "the long video is done but no energy to cut short clips," and this kind of tool solves exactly the capacity problem, and Chinese-subtitle accuracy has become quite usable in recent years.
Main features
- Automatic highlight-segment extraction from long videos
- Automatic vertical short-clip cropping
- Automatic subtitle generation
- Speaker smart auto-focus
- Multi-platform format output
Common uses
- Podcast highlight clips
- Online-course marketing segments
- Reusing lecture recordings
- Social-content mass production
Key Features
- Automatic highlight-segment extraction from long videos
- Automatic vertical short-clip cropping
- Automatic subtitle generation
- Speaker smart auto-focus
- Multi-platform format output
Pros
- Cuts short-clip time from hours to minutes
- Smart auto-focus handles multi-person conversations well
- A free plan to test
Cons
- AI-picked segments may not match the creator's judgment
- Chinese subtitles still need manual proper-noun proofreading
- The free plan has a watermark and duration limit
Use Cases
- Podcast highlight clips
- Online-course marketing segments
- Reusing lecture recordings
- Social-content mass production
Editor's Note
Content reuse is creators' most practical leverage in 2026. A one-hour interview, cut well, can feed three weeks of social — the key is that the cutting itself can't eat up another three hours of your time.
FAQ
Are the AI-picked segments accurate?
About 70% usable. It's good at catching excited tones, laughter, and clear turning points, but can't catch contextual judgment like 'this sentence is especially meaningful to my audience,' so you still need to review once yourself.
How accurate are the Chinese subtitles?
Everyday conversation is quite usable, but proper nouns, mixed-in English, and names of people and places have a higher error rate, so a quick proofread before publishing is advised.
Related AI Tools
KOL Radar
Taiwan’s influencer marketing data platform that turns KOL selection from guesswork into numbers you can talk about.
雅婷逐字稿
Taiwan AI Labs' transcription service built for mixed Mandarin, English, and Taiwanese speech.
Runway
Professional‑grade AI video creation and editing—directorial vision beats prompts.