Last week, we both grimaced at the camera
Last week I was helping my friend Wang Fang in Shenzhen tweak her Xiaohongshu (Little Red Book) talking-head scripts. By take 8 her voice was shot. The two of us sat in front of the camera and laughed bitterly — this whole ritual felt deeply anti-human.
I've been stuck here too. Last year when I was recording a product intro, I spent an entire afternoon on a 90-second talking-head clip and ended up picking the worst take. It took me a while to realize: it's not that I'm unprepared — it's that speaking to a camera shouldn't be this exhausting in the first place.
Here's what Meta just shipped
At Meta's Connect conference a few days ago, they upgraded their AI assistant Muse with a feature called Muse Realtime Avatar — your digital twin can speak, with matching lip-sync, expressions, and voice.
The kicker: Meta ran blind tests with judges, and their avatars looked more natural than Runway and HeyGen. Those two are the most popular avatar tools abroad; in China, plenty of people use HeyGen or CapCut's built-in digital human.
Wang Fang saw it and said: "If this had existed earlier, I wouldn't have to record every day." Right now she posts 5 Xiaohongshu videos a week, and she has to record each one at least 20 times to find a usable take.
But Meta's Muse isn't accessible in China right now — you need a VPN. Still, it signals something: making content in the future really doesn't require showing your face.
What you can actually use today
Money: CapCut's built-in "Digital Human" feature is fully free; HeyGen's paid plan runs about ¥100/month and supports custom avatars; Runway starts around ¥200/month.
Time: From signup to your first video — CapCut ~30 minutes, HeyGen ~2 hours.
Tech barrier: All drag-and-drop. If you can post to WeChat Moments, you can use these.
First step: Open CapCut → bottom menu "Text" → "Digital Human" → pick an avatar → paste your script → generate.
I made one mistake too: the first time I fed my avatar an 800-character long script, it read it in one go and the audience couldn't absorb anything. Cutting each segment to under 60 characters made the talking-head output actually watchable.
Playbooks by stage
If you're just starting out and haven't begun making content yet: I'd suggest spending a month testing the waters with CapCut's free version. If you can stick with daily posting, then think about upgrading.
If you already have 1-2 clients: I'd suggest trying HeyGen or Runway and commissioning a custom avatar. Brand recognition matters at this stage.
If you're scaling up (team of 3-5): I'd suggest mapping out the workflow first — who writes copy, who reviews, who publishes. Tools are just amplifiers; digital humans work best for "batch-producing videos once the script is locked."
One last thing: this tool isn't for everyone. If you enjoy the process of recording on-camera, avatars will feel like a cage. Figure out what you actually want first.