How to Create and Publish a Complete Faceless YouTube Video
It's 2 a.m. You can't sleep. You're scrolling through your phone, thumb moving on its own, when suddenly a video thumbnail stops you. No face. No camera. Just simple graphics. Seven million views. You check the channel—24 videos, 164,000 subscribers, first video posted just three months ago. You sit up in bed. If they can do it, why can't you? You open your laptop and hit a wall immediately. You don't know how to write a script people will actually watch. Every tool that could help asks for a credit card and money. You close the laptop. That channel just gained another 100,000 views. You're still at zero.
This is the exact feeling of being stuck at the starting line for six months. By the end of this guide, you will have published a complete video—script, voiceover, every image, thumbnail, and title—for zero cost, without a camera, without design skills. This is not a "how to get rich" video. This is for the person starting from absolute zero tonight.
The Critical First Step Most People Get Wrong
Most people approach this process backward, and that single mistake is why their videos feel off and nobody can explain why. The fix starts with NotebookLM. Open the platform and load a pre-configured project containing the best-performing "stick figure" style videos currently dominating the platform. The goal is not to copy them—it's to extract the writing style and idea structures that make them go viral.
Open the chat and type "Stage 1." Send it. NotebookLM will read through the details of these viral videos, analyze how they're written, and identify the hooks they use to capture and retain attention. This analysis is what informs everything that follows.
Select a topic from the options NotebookLM presents. Any topic you prefer works. Copy the chosen topic.
Script Generation With Claude
Open a new tab and go to Claude.ai. Sign in with your Google account, accept the terms, and stay strictly on the free plan. Skip any upgrade prompts. Navigate to Settings, then Custom Skills, then Add Skill, then Upload Skill. Select the prepared skill file—the link is in the description. Type "Try" followed by a slash and select the skill you just uploaded. Paste the topic you extracted from NotebookLM.
Claude will generate a complete script, typically five to ten minutes in length depending on the topic.
Why use NotebookLM for topic discovery and Claude for scripting rather than doing everything in one tool? NotebookLM searches YouTube, reads competitor content, analyzes what's working, and extracts the viral patterns. Claude can give you ideas, but NotebookLM wins decisively for competitive research. The combination of both produces a script built on proven patterns.
Arabic Translation and Voiceover
Copy the full script. Open a new tab and go to Gemini. Gemini is far stronger in Arabic than ChatGPT or Claude, which are notably weak in Arabic language tasks. Open a new chat and enable Flash Extended. Paste the translation instruction prompt and the full script. Gemini will return the script divided into paragraphs, translated into clear, natural Arabic.
The paragraph division is intentional. Each paragraph is kept to approximately 700 characters maximum.
Go to ElevenLabs. Sign up with your Google account. Skip to the free plan. ElevenLabs provides 10,000 free credits. If you exhaust them, simply register with a different Gmail address.
Navigate to Voices, select Arabic, and choose the Abdullah voice. Add it to your library. Go to Text-to-Speech, select Abdullah, and leave all settings at their defaults.
Return to Gemini. Copy the first paragraph. Paste it into ElevenLabs and click Generate. The Arabic voice quality is exceptional—currently the best Arabic text-to-speech available. Download the file and name it sequentially. Repeat for every paragraph.
Why process paragraph by paragraph? On the free plan, if you submit a long text block, the voice will change mid-generation and won't maintain the same character consistency. The tone at the beginning will differ noticeably from the end. Processing in small chunks preserves the Abdullah voice throughout.
Transcription for Image Prompts
After converting the entire script to audio, open CapCut. Start a new project. Import all audio files in order from last to first. Normalize loudness if the option is available (optional). Export the complete audio track as a single file.
Return to Gemini. Open a new chat with Flash Extended enabled. Paste the transcription prompt and attach the exported audio file. Gemini will generate a complete transcript with timestamps. The accuracy may not be 100%, but that's acceptable for the next step.
Copy the full transcript. Return to Claude in the same chat. Type "Stage 3" and paste the transcript. Claude will generate complete image prompts for every three-second segment of the video, formatted and ready for batch processing. When it finishes, type "Yes" and Claude will compile all prompts into a single TXT file for download.
Batch Image Generation
Go to Google AI Studio and open Google Labs. Create a new project with Google Flow. Accept the terms and continue.
Instead of manually generating each image one by one—which would take hours—install the ZapyFlow Chrome extension. Add it to your browser. This tool accepts a text file of prompts and automatically generates images for every segment without manual intervention.
In ZapyFlow, upload the TXT file Claude produced. Ensure Agent Mode is off. Select Nano Banana 2 as the model—it's free. Set the aspect ratio to 16:9. Set the output folder name. Click Start Generation. The tool will process all prompts automatically, downloading images to your computer as it works. If any individual prompt fails, copy it and generate that image manually. The result: 185 images generated without manual effort.
Video Assembly
Return to the CapCut project where the audio is already placed. Import all generated images. Drag them all to the timeline. The images will extend beyond the audio length, so select all images, scroll to the end, and trim to match the final frame.
Starting from the beginning of the video, align each image with its corresponding audio segment. The prompts were designed per three-second interval, so most images will naturally match their audio.
For segments longer than three seconds, add simple animations to prevent static frames. Go to Google AI Studio's video generation, select the Omni Flash model, set duration to six seconds, and prompt for subtle animation. Add these animated clips to break up the visual flow. This technique only needs to be applied to the first minute; extending it throughout the entire video produces even better results.
Add transitions sparingly—a black fade distributed across different points in the timeline, not between every image. Add subtle sound effects like TV static lines if desired. Export at maximum quality.
Title, Description, Tags, and Thumbnail
In the same Claude chat, type "Stage 4." Claude will generate a strong, viral-optimized title based on the script, along with a complete description and relevant tags.
Copy the tags into the YouTube upload form. Tags matter significantly for discovery. Select the video language and create a new playlist if desired.
For the thumbnail, ask Claude to generate five thumbnail prompts designed to go viral and grab attention. Copy the best prompt. Use ZapyFlow with the same settings to generate the thumbnail images. Select the strongest result and upload it.
Schedule the video for publication.
What Just Happened
You created a complete video—script, voiceover, every image, thumbnail, and title—from a blank page. Zero cost. No camera. No face on screen. No design skills. The person scrolling at 2 a.m. watching others succeed while stuck at zero was you at the beginning of this process. That person no longer exists.
"Later" is the graveyard where ideas die. Make one video tonight. Publish it. The prompt file is pinned in the first comment. Every tool mentioned is in the description, all completely free. The next video in this series covers how to grow the channel from its first upload into something sustainable. The only thing standing between you and your first published video is one blank page. Fill it.
