How to Create UGC Videos with AI in 60 Seconds
UGC converts better than studio content because it feels like a real person talking to camera. The problem has always been production: $150–500 per creator video, a week of turnaround, and briefs that come back wrong. AgentMedia collapses all of that into one command — you describe the actor and the line, it returns a realistic, ready-to-post 9:16 video. No photo, no actors, no editor.
How the selfie generator works
The selfie generator is AgentMedia's core product. From a text description it builds the entire video in three stages:
- 1Actor generation. A realistic person is generated from your one-line description, with a character sheet that locks their look.
- 2Photographic storyboard. A wireframe board plans the framing, gesture, and scene action so the delivery matches your direction.
- 3Video render. The final 9:16 video is generated at 5, 10, or 15 seconds — natural motion, expressions, and delivery.
You never pick a model or touch a timeline. Script and direction in, finished video out.
Step 1: Describe your actor
One sentence is enough. Age, look, energy — whatever matters for your audience. No photo upload, no filming, no licensing conversation.
"25yo asian woman, long wavy dark hair, soft smile"
Running a campaign? Persist the actor once as a character and every video stays on-model: agent-media character returns a char_xxxxxxxxxx id you can reuse in every generation — same face, every variant.
Step 2: Write the script and the scene
Two inputs: what the actor says, and what is happening in the frame. The scene direction is what makes the output feel like real UGC rather than a floating talking head.
script: "I keep getting DMs about my hair oil routine"
scene: "standing by a bright vanity, showing a small amber hair-oil bottle and scrunching one curl mid-line"
Keep scripts conversational and under ~30 words for a 10-second video. Hooks that work on TikTok work here: start mid-story, name something specific, skip the brand intro.
Step 3: Run one command
$ agent-media selfie \
--description "25yo asian woman, long wavy dark hair, soft smile" \
--script "I keep getting DMs about my hair oil routine" \
--scene-action "standing by a bright vanity, showing a small amber hair-oil bottle" \
--duration 10
Generating actor + character sheet...
Planning storyboard...
Rendering 9:16 video (10s)...
✓ ugc-final.mp4 — ready to post
The same generation runs from wherever you work: the web app, the REST API, the TypeScript or Python SDK — or inside Claude, Cursor, and n8n via the MCP server. Output is 1080p vertical, no watermark.
Step 4: Caption it and publish
Most viewers watch with sound off, so captions carry the message. The subtitle generator burns styled, word-synced captions onto any video — it transcribes automatically, or you can pass your own transcript:
$ agent-media subtitle ugc-final.mp4
$ agent-media publish --channels tiktok,instagram,youtube
Publishing is built in — schedule it or fire-and-forget to TikTok, Instagram, YouTube, and X. Your agent can run the whole loop: write the caption, generate the video, post it.
What it costs
Plans are credit-based and start at $39/month (Creator, 3,900 credits); a 10-second selfie video costs a few hundred credits, so a single subscription covers a steady posting schedule that would cost thousands of dollars with hired creators. Exact per-generator pricing is on the pricing page. Videos come in 5, 10, and 15-second durations — and the fastest way to find your winning ad is to generate five hooks and let the algorithm tell you which one works.
Ready to create your first UGC video?
One description, one command, one published video — in about a minute.