The agent writes.
You perform.

PuppetAgent is an AI agent with a human as one of its peripherals. The agent picks the topic, writes the script, and delivers it to your phone. You look into the camera and speak.

"BIGVU is a teleprompter with an AI script button.
PuppetAgent is an AI pipeline with a human as a recording device."

Script loaded
The future of content
creation isn't an avatar.
It's you. Directed by an agent.
Script delivered in under 5 seconds.
Voice scrolls as you speak.
REC
PuppetAgent Telegram Paste
POST /api/jobs/script
→ status: queued
The shift

The human is no longer
the operator.

Every other tool builds a workflow around a human who initiates. PuppetAgent inverts it.

Every other teleprompter
Human opens the app
Human writes or imports a script
Human presses record
Human edits and exports
Human uploads to a service
PuppetAgent
🤖 Agent picks topic from news or trends
🤖 Agent writes a 40-line script
📱 Script arrives on your phone automatically
🎙️ You read it — script follows your voice
🤖 Agent processes and delivers the final video

The human handles one thing: look into the camera and speak.
The agent handles everything before and everything after.

How it works

Five steps. One of them is yours.

The rest runs without you.

01
Agent

Agent writes the script

Your OpenClaw agent (or any script) picks a topic, generates 40–50 lines, and POSTs it to your PuppetAgent Server. One HTTP call. No SDK required.

02
Automatic

Script arrives on your phone

The app polls your server every 5 seconds. When a script is queued, it loads instantly — no notification, no tap required. Pick up your phone and it's ready.

03
You

Look into the camera and speak

The current line stays pinned to the top of the screen — close to the selfie camera. Vosk offline speech recognition scrolls to the next line as you finish each one. No internet required for STT.

04
Automatic

Recording uploads when you stop

Hit stop. The app uploads the mp4 to your PuppetAgent Server automatically. You're done. The pipeline takes it from here.

05
Pipeline

Post-processing delivers the final video

Voice cloning, background replacement, captions, gaze correction, lip sync — all running on your own hardware. You review a preview before the final render commits. Then it's in your Telegram.

Any agent. Any language.
curl is enough.

PuppetAgent Server exposes a plain REST API. No SDK. No auth dance. If your agent can make an HTTP POST, it can direct a human performance.

The performer's phone picks it up within 5 seconds. Your agent doesn't wait — it fires and moves on.

ClawHub skill available — one install, any OpenClaw agent
# Push a script to the performer's phone

$ curl -X POST https://your-server/api/jobs/script \
  -H "Content-Type: application/json" \
  -d '{"title":"Today",
     "script":"Line one.\nLine two.\nLine three.",
     "auto_start":false}'

# Response
{"job_id":"chr-a2b6f2b0","status":"queued"}

# Phone picks it up within 5 seconds. That's it.
Why it's different

Built for the pipeline,
not the human.

Every design decision optimizes for agent-driven workflows, not manual use.

🎙️

Offline voice auto-scroll

Vosk runs entirely on-device. No cloud. No cost per minute. No latency. No internet required for the STT. Works in airplane mode. PromptSmart does on-device too — but iOS only, proprietary engine. No Android app uses Vosk.

📡

HTTP script queue

Any system that can POST JSON can deliver a script to the performer's phone. No other teleprompter app exposes this interface. It's like a print queue for human performance — agent pushes, human executes.

🧑

Real human. Not an avatar.

Every other OpenClaw video pipeline uses AI avatars. PuppetAgent is the only one that routes through a real human on camera. The trust, micro-expressions, and authenticity that AI can't replicate.

🔓

Free. No subscription. Ever.

Apache 2.0 on Play Store. Self-host the server on any machine. No account required to use the app. No cloud STT means no usage bill. The entire market charges $9–30/month. This costs nothing.

The competition

Every other app assumes
you're the one in charge.

Only PuppetAgent is built for the case where you're not.

Feature BIGVU PromptSmart+ Teleprompter.com PuppetAgent
Voice auto-scroll Cloud STT On-device, iOS only Cloud STT ✓ Vosk, offline, Android
Records video
Offline STT ✗ Cloud only ✓ iOS, proprietary ✗ Cloud only ✓ Android, open source
Agent HTTP push API ✓ Full REST queue
Upload to your own server
No subscription required ✗ From $19/mo ✗ $10/mo ✗ Subscription ✓ Free forever
Open source ✓ Apache 2.0
Open source

Your voice. Your scripts.
Your server. Your data.

No cloud STT means your voice never leaves your device during recording. No account means no profile. Self-hosted server means your scripts and videos stay on hardware you control.

Android app — Apache 2.0 Server — AGPL-3.0
View on GitHub 🎭 ClawHub Skill

Look into the camera.
The agent handles everything else.

Free on Google Play. Apache 2.0. Self-host the server in 5 minutes.

Also available as a ClawHub skill for OpenClaw agents · Server source on GitHub