Been testing Opus 5.5 to rebuild my ENTIRE Faceless AI Avatar production workflow…
And gotta say this is lowkey insane.
So here’s an 8-step prompting protocol I’m using to scale 6-figure channels with it:
(Abusing the fact it’s cheaper, faster, and better than Fable 5.1…)
Prompt 1: the channel brain (run once per channel)
Create a Claude Project. Upload your 10 best competitor transcripts, your analytics export, your top 20 outlier titles and your thumbnail references. Then:
"Act as the head of production for this faceless YouTube channel. Read everything uploaded and build a channel bible covering: target viewer (age, country, device, when they watch), the emotional promise of every video, recurring story structures, title formulas, thumbnail rules, narrator tone, pacing rules and banned phrases. Include: no em dashes, no rhetorical questions as transitions, no 'welcome back to the channel'. Summarise it, then ask me what's missing before we write anything."
Prompt 2: outlier hunting (connect NexLev MCP first)
"Use NexLev MCP to find channels in [NICHE] under 6 months old with at least 3 videos above 100k views. For each, pull the 5 most recent and 5 best uploads. Flag channels where the best 5 are more than 5x the recent 5 (viral-spike traps, skip these). Return the survivors with estimated RPM, upload frequency and 48-hour view velocity on their last 3 uploads."
Prompt 3: packaging
"Here are 40 outlier titles from this niche. Reverse-engineer the psychological mechanism behind the clicks, not the surface pattern. Then write 25 titles for [TOPIC] built on those mechanisms, rank them by predicted CTR, and give me 3-5 words of thumbnail text for the top 5. Every thumbnail text needs an emotional verb."
Prompt 4: research brief (where Opus 5.5 earns its money)
"Research [TOPIC] and build a fact brief before any script gets written. Verify every date, name, location and quote against at least 2 sources. Flag anything where sources conflict. Mark anything you couldn't verify as UNVERIFIED instead of guessing. Then list the 5 most dramatic moments in chronological order with the source for each."
Prompt 5a: AI story script
"Using the fact brief and the channel bible, write a [LENGTH]-minute narration script. Cold open with a specific date, place and character in the first 15 seconds. 10 chapters. Cliffhangers at the end of chapters 3, 5, 7 and 9. Alternate high-tension scenes (3 minutes max) with slower character moments. The payoff lands at roughly 75% of runtime. The final chapter connects the story to something bigger. Follow every banned-phrase rule in the bible."
Prompt 5b: AI avatar script
"Rewrite the same story for an on-camera AI presenter. First person, direct to camera, sentences under 18 words so the lip sync stays natural. Mark every B-roll cutaway as [BROLL: description] and every emphasis beat as [PAUSE]. Keep the presenter on screen no more than 40 seconds at a time."
Prompt 6: visuals (connect Higgsfield MCP)
"Read the script and write one image prompt per scene change, around 18 total. Lock each character's age, face, clothing and hair across every prompt. Generate them through Higgsfield with GPT Images 2.0 in 16:9. Then pick the 3 most dramatic scenes and animate them with Seedance 2.5 at 8 seconds each for the intro and key beats."
Prompt 7: the edit list
"Turn the script and the generated assets into a timestamped edit decision list for CapCut. For every line: which asset is on screen, when to cut, where the slow zoom goes, where the animated clips sit, and which words the captions should emphasise. Output as a table I follow top to bottom."
Your editor (or you) stops making decisions and starts following instructions. A 1-hour video assembles in under 30 minutes.
You can also try using CapCut MCP for this, but it’s pretty ass ngl.
Prompt 8: the weekly audit
Export your YouTube Analytics as a CSV and drop it in:
"Find the 3 variables most correlated with videos beating my channel average. Look at publish day and time, title structure, thumbnail style, video length, topic, and the timestamp where retention dropped hardest. Then give me next week's 5 video ideas built on whatever's winning, with titles."
Run the text side of all 8 prompts through Claude and you're looking at roughly $1-2 of Opus 5.5 per finished video.
Opus 5.5 is cracked at reasoning, avoids stupid data mistakes, and sounds like a human for less than Fable 5.1, which you’ll notice immediately.
All PERFECT for avoiding shadowbans, demonetizations, and being attacked for spreading misinformation online - killing your channel trust - which would happen often if using cheaper models more prone to hallucination.
You can get all skill files and my 1:1 help to implement this btw if you join Elevate.
DM me “UNO” for details.
The first 6 seconds of a faceless YouTube video determine whether the algorithm pushes it to 1,000 impressions or 100,000.
And 90% of AI story channels are failing this test with static images that trigger the "slideshow" signal in the viewer's brain before they consciously decide to stay or leave.
One production change fixes this entirely…
Three animated clips at the start of the video. 6 seconds each. Generated through Seedance 2.5 on Higgsfield MCP. Total generation time: under 5 minutes. Total production cost: near zero.
The mechanism is neurological, not aesthetic.
When a viewer clicks on a YouTube video, the first few seconds activate a snap judgment: is this produced content worth watching, or is it a slideshow I should scroll past? Movement in the opening frames triggers the "produced content" classification. Static images trigger the "slideshow" classification.
The algorithm reads this through the retention curve. Videos with strong 6-second retention get tested against broader audiences. Videos where viewers bounce in 3-5 seconds get deprioritized before the content had a chance.
The 3 animated intro clips are not an aesthetic upgrade. They're a retention mechanism that changes how both the viewer's brain and the algorithm's scoring system evaluate the video.
Here's the prompt:
"Use Higgsfield MCP to generate 3 animated video clips using Seedance 2.5. Each clip should be 6 seconds, 16:9 aspect ratio. Build each clip from the hook of the script: [describe opening scene with character positioning, lighting, camera angle, atmospheric mood]. Output at highest available resolution."
The 3 clips go at the front of the timeline in CapCut. Voiceover starts over the animated clips so the viewer hears the hook while watching movement. By the time the content transitions to static images at 20 seconds, the viewer has committed.
The rest of the video uses 15 static images with subtle zoom effects. The heavy lifting is done in the first 6 seconds.
This is the single lowest-effort, highest-impact change most faceless creators haven't made. Costs nothing through Elevate's tools. Takes 5 minutes.
Measurable retention improvement within 3-5 videos.
The full production system with Seedance 2.5 setup and TubeChef for automated editing is inside Elevate at elevate(.uno). DM me if you want to see the before-and-after retention data
I've tested over 300 videos across 10s of AI story channels.
These are the hooks, formats, emotional words, niches, angles, and titles generating the most views and revenue right now.
(Save this and come back to these formulas every time you plan a new video.)
Winning niches (with RPM ranges):
Romance / Alpha King stories run $15 to $35 RPM with an audience of women 55+ in USA and UK. Massive demand, 20+ channels pulling 1k+ views.
Military history and war documentaries run $18 to $30 RPM with an audience of men 35-65 in USA, UK, and AUS. Outlier potential is insane (1M+ views from 25k-sub channels).
True crime and unsolved mysteries run $12 to $25 RPM with a mixed audience 25-55. HIGH volume niche, easy to find stories from court transcripts and cold cases.
Historical empires and ancient civilizations run $15 to $28 RPM with an audience of men 30-60. DEEP content library covering 1,000s of untouched stories across every civilization.
Survival and wilderness stories run $10 to $22 RPM with seasonality (summer dips, winter spikes) and a strong suggested video loop.
Horror and creepypasta run $8 to $18 RPM. Lower RPM with HUGE volume and viral ceiling. Younger audience entering.
Winning title formulas:
"The [adjective] [trap/strategy/mistake] [authority figure/group] [action verb]" turns into "the genius trap every military strategist knows."
"How [nation/person] [emotional verb] [nation/person]" turns into "how America fooled Britain."
"[person] was [injustice/crisis], [unexpected twist]" turns into "she was left to freeze at the castle gate, the Alpha King carried her in."
"The [time period] [person/group] [surprising action]" turns into "the 1781 farmers who destroyed Britain's best soldiers."
"Why [big entity] [failure verb] against [underdog]" turns into "why Rome's greatest legion lost to german farmers."
The pattern: specific plus emotional plus incomplete information. Every winning title gives the viewer ENOUGH information to understand the topic while leaving them short of satisfied without clicking.
Winning hook structures (first 30 seconds of script):
The date drop: "on [specific date], [specific person] [dramatic action] in [specific location]." Grounding the story in a precise date and place triggers the viewer's documentary-brain. The specificity signals "this happened, this is real, stay and learn."
The underdog inversion: "[weak group] faced [strong group] and everyone expected [obvious outcome]. What happened next is studied at [prestigious institution] to this day." The gap between expectation and reality IS the hook. The viewer stays to close the gap.
The character flaw reveal: "[famous figure] had one weakness [specific flaw]. And on [date], the weakness almost destroyed everything." Humanizing a historical figure through a specific flaw makes the viewer care about the person far beyond the topic.
The countdown: "[number] [time unit] is all [person] had before [consequence]. Here's how they used every single one." A ticking clock creates urgency the viewer inherits.
Winning emotional words for titles and scripts:
High-click words: genius, fooled, tricked, crushed, humiliated, destroyed, betrayed, abandoned, stolen, forbidden, secret, buried, forgotten, impossible, deadly, brutal.
High-retention words inside scripts: "what nobody knew was..." / "and this is where everything went wrong" / "but [person] had already planned for this" / "the next [time period] would change everything."
Words to avoid: interesting, amazing, incredible, awesome, shocking (overused, low-trust, feels clickbaity without substance).
Winning video formats:
Single-event deep dive: one battle, one crime, one scandal. 20-45 minutes. The ENTIRE video is about one event told in extreme detail. This is where outliers live.
Biographical arc: one person's rise, reign, and fall. 30-60 minutes. Works best with figures who had dramatic turning points.
Mystery/investigation: unsolved case or cold case. 25-40 minutes. The "who did this" question drives retention through the entire video.
Versus/comparison: nation vs nation, leader vs leader, army vs army. 20-35 minutes. The conflict frame creates natural tension without needing to manufacture drama.
Winning video angles (the twist 9 out of 10 creators miss):
The angle is the FRAME around the topic rather than the topic itself.
Same topic, different angles.
Topic: battle of Cowpens.
Angle 1: "the genius trap every military strategist knows" (strategy frame).
Angle 2: "the farmer who humiliated Britain's best officer" (underdog frame).
Angle 3: "the 60 minutes where America won the revolution" (stakes frame).
Each angle pulls a different audience. The strategy frame attracts military enthusiasts. The underdog frame attracts general viewers who love comeback stories. The stakes frame attracts history buffs interested in turning points.
The outlier performers on YouTube almost always use the BROADEST angle. "The genius trap" outperforms "the battle of Cowpens" because the first title appeals to anyone curious about strategy. The second speaks only to American revolution fans.
Rule: abstract the specific topic into a universal curiosity trigger. Then let the specifics live inside the video.
What to do with all of this:
Pick ONE niche from the list. Validate with 10+ channels pulling 1k+ views. Select the title formula and hook structure matching the niche's top outliers. Produce 3-5 videos per week using the winning format. Iterate based on CTR and APV data.
Every formula here came from testing rather than guessing. The data is the data.
Full breakdowns and templates for each formula inside elevate(.)uno If you want the complete system.
If you're getting 0 views on your faceless YouTube channel, the algorithm doesn't "hate you" bro
Your IP is cooked. Let me explain how to fix the entire setup in under 10 minutes.
I burned 14 channels before figuring this out. Uploaded produced content for weeks. 0 impressions. 0 browse features. 0 explanation from YouTube.
The content wasn't the problem. The infrastructure underneath was broken the entire time.
Here's the exact setup:
Step 1: download Dolphin Anty.
This is the anti-detect browser I've been using for a long time and the thing works flawlessly. You get 5 free profiles so you don't pay until you're running 6+ channels. By that point, paying for the tool isn't an issue.
Step 2: proxies.
2 companies I recommend (not paid promos, these are what i personally use):
> Web share - unlimited data or 5,000 GB plan. Private IPs nobody else is using. 25 IPs for ~$30. 1 IP = 1 channel. Never change the proxy on a channel once assigned.
> Tube proxies - 100% tailored to YouTube. Slightly pricier but they guarantee 0 fraud score on every IP. Only downside is stock runs out fast.
Never use shared IPs. Never use the free proxy tiers. those carry mid-to-high fraud scores and you'll burn the channel before uploading a single video.
Step 3: create the profile.
Open Dolphin. Name the profile (1 profile per channel). Add your proxy in the proxy section. Click "new fingerprint" and make SURE the hardware data makes sense for your actual device.
Running an M2 MacBook? Set the renderer to M2. The more recent and consistent the hardware profile, the higher the trust score.
Do NOT set apple silicon as the processor with an nvidia GPU as the renderer. That combination doesn't exist on any physical machine. YouTube sees the mismatch, flags the profile as synthetic, and kills your reach. I burned a fully validated channel this way.
Step 4: verify before uploading.
Open the Dolphin profile. Go to browserleaks(.com). check the fraud score. If the score is high or the fingerprint looks off, swap to a different IP from your plan.
30 seconds. Saves you weeks on a flagged profile.
Rules after setup:
Never create a channel inside the anti-detect browser. Log into a channel you previously created or an aged channel. 99% of the time you won't even be able to create 1 inside these browsers.
Never access the channel through YouTube Studio on your phone. Always log in through dolphin with the same proxy. The studio app links all your channels together through 1 device fingerprint and undoes the entire separation.
This setup is the difference between 0 views and 800+ views on the same video.
Full walkthrough video on my YouTube if you want the screen recording version.
It's completely legal to:
• Go on google trends
• Find problem people are struggling with
• Turn those problems into digital assets
• Without revealing your name
• Without showing your face
I made $65k last month doing it.
Here's exactly how it works:
THIS IS F**KING GOLD.
This guy just REVEALED how to get paid $19,772 from YouTube EVERY month
In just 20 seconds, he shows the workflow, step by step. Totally free.
Idea → Create → Monetize
More details in the article below
Save this, you’ll thank me later
There's a method for stealing traffic from any large YouTube channel within 48 hours of them posting and the algorithm does the work for you
The people using it don't talk about it because the moment it gets popular, everyone floods the window & it stops printing
Here's how it works:
When a 500K+ subscriber channel posts a video about a topic & gets 2 million views, YouTube's recommendation engine immediately starts hunting for MORE content on that topic to feed the same audience
The sidebar. The "up next" queue. Homepage recommendations. YouTube actively looks for related videos to fill those slots because 2 million viewers just told the algorithm they care about this subject
If you publish a video on the same topic within 48 hours with similar keywords in your title → YouTube will test YOUR video in those recommendation slots. Your 5K-subscriber channel shows up next to a 500K-subscriber channel's video because the algorithm matches on topic, not authority
The 48-hour window matters. YouTube's recommendation system runs hot on fresh topics. After a week the slots fill up & the topic cools. First mover advantage is real & measurable
The method:
→ Set Google Alerts for the 5-10 biggest channels in your niche
→ When they post, you have 48 hours
→ Make a video on the same topic ~ different angle. Agree, disagree, go deeper, apply it to a specific industry
→ Use similar keywords in the title, not identical
→ Upload within the window
A channel I built content strategy for went from 4K → 22K subscribers in 3 months doing nothing but response videos within the 48-hour window. Every time a big channel posted, he published a deeper take the next day. YouTube kept feeding him their audience
The big channel never found out. The algorithm doesn't notify creators when their videos generate recommendations for competitors
Their $50K production budget validated the topic. Your $0 production budget captured the spillover. They did the market research. You did the follow-through
Some people call this piggybacking. I call it reading the room
(link in bio)
There is one filter that separates faceless YouTube channels printing $10k/mo from the ones stuck at zero views for 3 months straight.
And the filter has nothing to do with content quality, thumbnail design, or scripting ability…
It's the niche validation. And most people skip it entirely.
Here's the method. Takes 2-3 hours. Saves months.
Open YouTube in incognito mode. Search for channels in your target sub-niche. Find 10 separate channels pulling 1,000+ views per video consistently across their last 15-20 uploads.
Not one viral hit. Not a channel with 500k on their best video and 200 on everything else. Consistent output across 15-20 recent uploads.
If you find 10 channels meeting the threshold, the niche has real demand. The algorithm is pushing content in this space. New entrants are being rewarded. The audience exists and they're watching.
If you struggle to hit 5, walk away. The niche is either dead, dying, or so small the algorithm doesn't know what to do with the content. No amount of production quality, scripting technique, or upload consistency fixes a demand problem.
The reason this filter works is straightforward.
YouTube's algorithm needs data to recommend your content. When 50+ channels are producing content in a niche and the audience engagement is strong, the algorithm has a deep recommendation model to work from.
When you enter a niche with zero competitors, the algorithm has no data. No recommendation model. No audience profile to match your content against. Your uploads sit in a vacuum.
The niches that feel "untapped" and "undiscovered" are almost always graveyards. Nobody is there because the audience isn't there. You're not early. You're alone.
I validated every niche I entered this way before uploading a single video. The channels where I skipped validation and went on gut instinct? Dead. Every single one.
If you want to speed this up, NexLev MCP connected to Claude does the same validation in one conversation. The manual incognito method works too. The discipline is what matters. Validate before you produce.
Everything behind the validation system, the niche finder, the full production course, free unlimited ElevenLabs, and coaching directly with me is inside Elevate at elevate(.uno).
DM me if you want help picking your first niche
The creators printing 6 figures/mo from faceless YouTube aren't doing so from 1 channel.
They're running 6, 8, 12 channels simultaneously and treating every single 1 as a disposable asset inside a larger system. Let me explain why this is the ONLY way to build a sustainable business on the platform.
I call the strategy the Hydra Method.
Named after the mythological creature where every time 1 head gets cut off, 3 new heads grow back.
Nearly every creator approaches YouTube like their entire business depends on 1 channel surviving forever. They spend months building and mentally attach themselves to the asset. the channel becomes their identity.
The problem? YouTube has never guaranteed permanence. channels get demonetized. Channels get terminated. Policies change without warning. Entire niches get hit with inauthentic content waves.
If your entire revenue depends on 1 channel surviving forever, you built a fragile business model.
The Hydra Method flips the whole thing:
For every channel the platform kills, you start 3 more. You leverage the collaboration feature to transfer algorithmic trust from your surviving channels to the new 1s. You reset your AdSense publisher ID to remove the "infected" association from demonetized channels.
I've pulled channels from 15 impressions per video to 400+ in 24 hours using 1 strategic collaboration.
The method only works when production costs stay low. 9 out of 10 people hear the concept and immediately want to apply the strategy to complex projects requiring researchers, editors, and $5k/mo production budgets per channel.
The content formats built for the Hydra are the formats producible at scale with near-$0 cost: AI stories, sleep content, simple slideshow formats.
When a channel goes down, replacing the asset doesn't require rebuilding a company. You replicate a process that already exists.
The 6-figure math through the Hydra lens:
1 channel: $10-$20k/mo. Vulnerable to 1 demonetization event.
3 channels: $30-$60k/mo. If 1 goes down, 66% survives while you launch replacements.
6 channels + 1 VA: $60-$120k/mo. If 2 go down simultaneously, 66% still runs while the Hydra grows back.
The full Hydra Method logistics, AdSense recovery system, and collaboration sequencing framework are inside elevate(.)uno
Realistically, you can earn up to a thousand dollars on this platform.
I shared the update in the morning.
Check me out on instagram https://t.co/lDVZjVlZyp
If you haven’t gone to zero once before in your life, you won’t understand what she’s saying.
Zero is a demonic place to be especially if you don’t have a helper.
Bro, I couldn’t afford eye-drop of five thousand Nara (5,000).
Touching millions then going back to $0 will have you questioning God.
I wept like mad.
This guy just LEAKED the EASIEST way to make money in 2026:
Claude Code + YouTube = 62,000 $/month
(In just 13 minutes)
Save it. You'll thank me later.
the tool that makes the videos, runs on your own machine:
https://t.co/7WkKW6j4Em
prompts and everything that doesn't make it into videos:
https://t.co/O2Rxfw9B4Q
every faceless youtube creator's instinct is to lead with their strongest point.
this instinct is wrong. and the data from 8,000+ scripts proves it. the channels with the highest completion rates put their BEST point in the MIDDLE, not the beginning. and the difference in retention is 15-25% on completion rate alone. let me show you the ordering that holds...
2nd best first. best in the middle. 3rd best last.
sounds counterintuitive. the data makes it obvious.
if you lead with your strongest point, the trajectory goes DOWN from there. each subsequent section feels weaker than the last, even if the information is equally valuable. the viewer's brain registers a declining trajectory and the subconscious response is "this video peaked early. i've gotten the best part."
when you lead with your 2nd strongest point, the trajectory goes UP. the viewer hits your best point at the midpoint and experiences an escalation. the script feels like it's getting BETTER as it progresses.
the 3rd best point closes the video on a "good" note that's still above where the viewer started. the ending doesn't need to be the strongest. it needs to feel satisfying after the midpoint peak.
the psychological principle behind this is called progressive revelation. the viewer's brain perceives an escalating pattern of value delivery. each section feels more valuable than the last (until the midpoint peak), which creates the perception of "this keeps getting better."
the perception of escalation is more important than the actual quality ranking.
a video where sections go 8/10, 10/10, 7/10 FEELS like it kept getting better because the peak happened late. a video where sections go 10/10, 8/10, 7/10 FEELS like it peaked early and declined, even though the average quality is identical.
i track this across every script i write. the ordering is consistent: 2nd-1st-3rd outperforms 1st-2nd-3rd by 15-25% on completion rate. same content, same hooks, same editing. the only variable is the ORDER the points are presented.
go through your last script. check whether your strongest section is first or in the middle. if it's first, swap it with your 2nd strongest. the retention graph responds within the next 3 uploads.
DM me "RETAIN" if you want FacelessOS to structure this ordering into your scripts automatically
haris
YOU NEED TO BE LOCATIONMAXXING
the same YouTube video, in the same niche, pays $11.95 per 1,000 views in the US and $0.42 in pakistan.
a 28x spread, and the only difference was which country got served the video.
the algorithm makes that call by reading what’s IN your script. almost nobody writes with this in mind.
say you want 10x more revenue. there are 2 ways to get there, and they are not equally hard.
option 1: get 10x more views, which means outcompeting the entire platform.
option 2: keep your views and move your audience from a $1 RPM country to a $12 one.
one of those takes years and might not work.
the other takes a few rewrites.
the beautiful thing about going after US viewers is that you get paid twice more after adsense: they watch on TVs (longer sessions), they return to videos, and they BUY things, courses, products, tools. the $0.42 viewer consumes. the $11.95 viewer transacts.
YouTube decides who sees your video by reading every signal in it. so you MUST condition every signal. Here are the 7 levers you control here:
> word choice. your food example is thanksgiving turkey, not tikka masala. small change, massive routing data
> cultural references. super bowl, not cricket world cup. each reference narrows the country before the algorithm reads your topic
> the examples you build videos around. a chess channel covering bobby fischer and magnus carlsen attracts a different map than one covering gukesh and arjun. same niche, different target demographic
> the voice. an american accent on your elevenlabs model pulls american viewers
> the avatar. if your channel has a character, the character's appearance is a demographic signal on its own
> the settings in your edits. show suburbs, diners, US highways. your b-roll is location data
> niche sophistication. finance and business framings pull higher-income audiences than vague curiosity-bait framings, even on the same underlying story
audit your last 5 scripts against that list and count how many signals point at a $1 country. i've never audited a low RPM channel where the answer wasn't sitting somewhere in these 7 levers.
your script was always the targeting.
now aim it.
if you want the conditioning checklist built into your script system, DM me "GEO" and i'll send it over
(200M+ views generated through AI scriptwriting skill files trained on 8,000+ scripts, one-time payment)
If you want to learn even more and get guidance and support from me and my team to help you build your own 6 figure YouTube channel, book a call with us:
https://t.co/lGMrFqg5lT