you + claude + youtube = $40K–90K per month
this guy racked up about 8,000,000 views in less than 2 months
without spending hours creating storylines, voiceovers, images, just him and AI
monetization earned him about $40K–90K
so in this article i will tell you:
how to make faceless stickman youtube videos with free ai tools
a complete workflow for turning a topic into a published long-form video, script, voiceover, visuals, edit, upload - using only free plans
no paid tools, no design skills, no team. these are the same mechanics behind doodle/stickman channels pulling millions of views off still images
the stack
everything here runs on a free tier:
the core principle: voiceover first
this is the rule the whole workflow is built on, so internalize it before touching any tool
most people build in this order: script → image prompts → generate scenes → record voice → sync it all together at the end.
that order is the problem
when visuals come first and voice comes second, you're forcing two things together that weren't built for each other
the result feels off, viewers can't name why, they just click away
flip it, voiceover first, scenes second, always
the reason: a finished voiceover has natural pauses between sentences, those pauses are your cut points
each scene starts and ends on a pause, so every cut is born from the rhythm of the voice instead of guessed at afterward
that rhythm, not the art, not the tools is what holds a viewer
every step below exists to serve this order
step 1 — write the script (claude)
this is your foundation; the cloud will serve as the connector between your steps
open claude on the free plan and give it this prompt:
prompt returns multiple viral angles, not one, pick the strongest
what to check: the script should sound spoken, not written
step 2 — generate the voiceover (elevenlabs)
in this step, we'll create the foundation of our video - the voice
which we'll use to structure the images and the video as a whole
go to https://elevenlabs.io/
step 3 — find the pauses (transcription)
in order to generate images for each part of the text and for claude to understand all parts of the video, you need to create timeframes for voice
to do this, we'll use https://foziscribe.ai/
with the free plan, you'll get 10 credits, which is enough for two 3 to 5 minutes transcripts in fast mode
go to the website, create an account, and in the “Transcribe” tab, upload your mp3 file with the audio and click “Transcribe Audio”
download the text file and send it to the chat with Claude where you previously generated the text
it will start creating prompts for all the frames you'll need
claude will generate prompts in batches, so tell it to keep going, and you'll receive a text file containing all the prompts once the process is complete
step 4 — generate the scenes (google flow)
open flow.google → new project → agent mode off → image mode on → 16:9 → one output per prompt → model set to nano banana 2 (strong results without burning through limits)
generating 100+ scenes by hand takes me like 40 mins, but you can use free bulk-generation chrome extension built for flow
do your own research before using it!
for convenience, i also renamed each photo to reflect the time it's used, that took about 30 minutes
once you've made sure that every photo turned out well, download the entire project from google flow (just regenerate any bad photos)
step 5 — edit to the rhythm
you'll need any video editor
for example, i used filmora
it's a paid program, so you'll either need to buy a subscription, figure out how to remove the watermark, or just use a different editor
import the voice track and the entire folder containing the materials
then place the images into their respective time frames, it’s simple
insert an image → move to the desired second → crop the image → insert the next one
step 6 — package and publish (claude)
back in claude, generate the metadata:
title, description, and tags tuned to your video.
then ask for five high-ctr thumbnail prompts, run them through google flow, and pick the strongest one
for example, this one is my favorite:
step 7 — publish (youtube)
upload to youtube, drop in the metadata, set the thumbnail, publish
common mistakes
result
ended up with a video like this, of course, still need to find the right editor or figure out how to remove the filmora watermark
so, i'll be waiting for your suggestions in the comments
