How I Built an AI Agent to Create Faceless YouTube Videos (No-Code)
Sabrina Romanov demonstrates a no-code automation workflow built in n8n that generates faceless AI YouTube/social media videos on autopilot. The system uses an AI agent to write scripts, a single API call to her platform Blotato to generate the full video, and then auto-publishes to multiple social platforms. The template is available for free download.
Summary
The video is a tutorial by Sabrina Romanov, an AI educator with over 700,000 followers, walking viewers through a no-code automation workflow that creates and publishes faceless AI videos to social media platforms automatically.
The workflow is divided into three main parts: writing the video script, creating the faceless video, and publishing to social media. The script-writing phase uses an AI agent node (powered by OpenAI GPT-4o) that brainstorms ideas based on a user-defined niche — the default being 'little known history facts about a famous person.' The agent selects an idea, researches it, writes a 15-second video script, and produces a video caption, all output in structured JSON format via a structured output parser.
The second phase relies on Blotato, a platform built by Romanov herself, which consolidates what would normally be multiple separate API calls (image generation, video generation, captions, voiceover, stitching) into a single API endpoint. The endpoint accepts parameters like template type (empty or POV), voice selection via ElevenLabs, caption position, animation settings, and the choice of text-to-image and image-to-video models. Available video models include FramePack (the default and cheapest), Runway, Luma, Kling 1.6, MiniMax, Hyen, and Google's V2. For images, Recraft is recommended for realistic human imagery, while Flux Pro is also highlighted. A wait node is used to allow rendering time, with a minimum of 10 minutes recommended when animating images into video clips.
The third phase handles publishing to social platforms including Instagram, TikTok, Facebook, YouTube, Threads, and LinkedIn. Users need to connect their social accounts in Blotato settings, copy their account IDs, and activate the relevant publishing nodes in the workflow.
Romanov walks through the minimal setup required — inserting a Blotato API key and one social account ID — and covers common failure points such as incorrect API key copying, insufficient wait times causing incomplete video retrieval, and account ID errors. She also discusses customization options including changing the niche, video duration, template style, voice, caption position, and animation settings. Support is available via an in-app messenger that escalates to Romanov herself within 48 hours.
Key Insights
- Romanov argues that most existing faceless video tutorials require separate API calls for image generation, video generation, captions, voiceover, and stitching — whereas her Blotato platform consolidates all of this into a single API endpoint, significantly simplifying the automation.
- Romanov states that animating images into video clips alone takes 5 minutes per render, and recommends a minimum wait time of 10 minutes in the workflow — longer for scripts exceeding 15 seconds — to avoid the 'get video' node returning a 'ready' status instead of an actual video URL.
- Romanov identifies FramePack as the default and cheapest image-to-video model, recommending it for 99% of use cases, while noting that Kling 1.6 and Runway are higher-quality but more expensive alternatives available within the same API.
- Romanov describes a POV-style template option within Blotato that interprets the input script and rewrites it into a POV-style video — a format she notes is currently going viral — as an alternative to the default 'empty' template which uses the script as-is.
- Romanov recommends using Recraft as the text-to-image model when the video requires realistic-looking images of people, citing it as superior to Flux Pro for that specific use case, while acknowledging Flux Pro as a strong general alternative.
Topics
Full transcript available for MurmurCast members
Sign Up to Access