VEED UGC

Generate UGC-style promotional videos with AI lip-sync. Takes an image (person with product from Morpheus/Ad-Ready) and a script (pure dialogue), creates a video of the person speaking. Uses ElevenLabs for voice synthesis.

Security Scan

The skill mostly does what it says (uploads an image + script to a ComfyDeploy endpoint and returns a lip-synced video), but its manifest and instructions are inconsistent about required credentials and it prints potentially sensitive API responses — review before use.

Purpose & Capability

The name and description (generate UGC videos with lip-sync/TTS) match the code and SKILL.md: the script uploads an image, queues a run at ComfyDeploy, polls, and downloads the result. Mentioning ElevenLabs is reasonable because a voice_id is used, but ElevenLabs credentials are not required by the bundled code (ComfyDeploy is the service actually contacted).

Instruction Scope

SKILL.md and the included script instruct the agent to upload user-supplied images and script text to https://api.comfydeploy.com and to include an Authorization Bearer API key. The code also prints debug info and the first 500 characters of API responses to stdout, which can leak sensitive data (including tokens or returned URLs). The skill will transmit user images and text to an external service — this is expected for the stated purpose but is a privacy/data-exfiltration consideration that must be explicit to users.

Install Mechanism

There is no install spec (instruction-only plus an included Python script). Nothing is downloaded from arbitrary URLs and no install-time code execution is requested. Risk from install mechanism is low.

Credentials

The skill manifest declares no required env vars, but the script requires a COMFY_DEPLOY_API_KEY either via --api-key or environment (COMFY_DEPLOY_API_KEY). That mismatch is incoherent and could confuse users. No ElevenLabs secret is requested (because ComfyDeploy handles TTS), which is plausible, but the SKILL.md's voice/ElevenLabs messaging could mislead users into thinking an ElevenLabs key is required. The script's debug printing of response bodies can expose credentials or other sensitive return values to logs.

Persistence & Privilege

The skill is not always-enabled and does not request persistence or modify other skill/system settings. It runs on-demand and does not escalate privileges.

What to consider before installing

Before installing or running this skill, note that: (1) the included Python script will upload any image and the script text you provide to https://api.comfydeploy.com — do not upload images of real people without explicit consent; (2) the script requires a ComfyDeploy API key (COMFY_DEPLOY_API_KEY or --api-key) even though the skill metadata does not list that requirement — this mismatch is suspicious and you should provide a key with minimal privileges; (3) the script prints API responses (first 500 chars) to stdout which can leak sensitive info to logs — review or remove these debug prints if you care about secrecy; (4) ElevenLabs is referenced only for voice IDs; no ElevenLabs credential is included, because TTS is performed by the ComfyDeploy workflow — verify this behavior with the service owner; (5) the source/homepage is unknown: prefer packages with a verifiable source or inspect the code thoroughly and run in an isolated environment. If you are unsure, do not use with real users' images or private scripts until you confirm the service/policy and fix the manifest mismatch (declare the COMFY_DEPLOY_API_KEY requirement).

Generate UGC (User Generated Content) style promotional videos with AI lip-sync using ComfyDeploy's Veed-UGC workflow.

Overview

Veed-UGC transforms static images into dynamic promotional videos:

1. Takes a photo of a person with a product (from Morpheus or Ad-Ready)
2. Receives a script (pure dialogue text)
3. Creates a lip-synced video of the person speaking the script

Perfect for creating authentic-feeling promotional content at scale.

API Details

Endpoint: https://api.comfydeploy.com/api/run/deployment/queue Deployment ID: 627c8fb5-1285-4074-a17c-ae54f8a5b5c6

Required Inputs

Input	Description	Example
image	URL of person+product image	Output from Morpheus/Ad-Ready
script	Pure dialogue text	"Hola che! Cómo anda todo por allá?"
voice_id	ElevenLabs voice ID	Default: PBi4M0xL4G7oVYxKgqww

⚠️ CRITICAL: Script Format

The script input must be PURE DIALOGUE ONLY:

✅ CORRECT:

Hola che! Cómo anda todo por allá? Mirá esto que acabo de probar, una locura total.

❌ WRONG - No annotations:

[Entusiasta] Hola che! (pausa) Cómo anda?

❌ WRONG - No tone directions:

Tono argentino informal: Hola che!

❌ WRONG - No stage directions:

*sonríe* Hola che! *levanta el producto*

❌ WRONG - No titles/labels:

ESCENA 1:
Hola che!

Just write exactly what the person should say. Nothing else.

Voice IDs (ElevenLabs)

Voice	ID	Description
Default	PBi4M0xL4G7oVYxKgqww	Main voice

More voices can be added from ElevenLabs

Usage

uv run ~/.clawdbot/skills/veed-ugc/scripts/generate.py \
  --image "https://example.com/person-with-product.png" \
  --script "Hola! Les quiero mostrar este producto increíble que acabo de probar." \
  --output "ugc-video.mp4"

With local image file:

uv run ~/.clawdbot/skills/veed-ugc/scripts/generate.py \
  --image "./morpheus-output.png" \
  --script "Mirá, yo antes no usaba esto pero ahora no puedo vivir sin él." \
  --voice-id "PBi4M0xL4G7oVYxKgqww" \
  --output "promo-video.mp4"

Direct API Call

const response = await fetch("https://api.comfydeploy.com/api/run/deployment/queue", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer YOUR_API_KEY"
  },
  body: JSON.stringify({
    "deployment_id": "627c8fb5-1285-4074-a17c-ae54f8a5b5c6",
    "inputs": {
      "image": "/* put your image url here */",
      "voice_id": "PBi4M0xL4G7oVYxKgqww",
      "script": "Hola che! Cómo anda todo por allá?"
    }
  })
});

Workflow Integration

Typical Pipeline

1. Generate image with Morpheus/Ad-Ready

   uv run morpheus... --output product-shot.png
   

2. Write the script (pure dialogue)

3. Create UGC video from the image

   uv run veed-ugc... --image product-shot.png --script "..." --output promo.mp4
   

Output

The workflow outputs an MP4 video file with:

• The original image animated with lip-sync
• AI-generated voiceover from the script
• Natural head movements and expressions

Notes

• Image should clearly show a person's face (frontal or 3/4 view works best)
• Script is spoken exactly as written - no interpretation
• Video length depends on script length
• Processing time: ~2-5 minutes depending on script length