Shade

𝗜𝗻𝘀𝗶𝗴𝗵𝘁𝘀 𝗼𝗻 𝗔𝗜, I𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗶𝘀 𝗯𝗲𝗶𝗻𝗴 𝗯𝘂𝗶𝗹𝘁 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄

[Shade @twoshados]

This guy built an AI pipeline that generates hyperrealistic fashion models in 47 minutes and now dropshippers pay him $1,400 to clone the entire system.

He got tired of watching e-com brands lose $8K per photoshoot when a single product angle changed so he built a 9-node workflow that generates 127 product videos from one Pinterest photo without hiring a single model.

Here's the exact breakdown:

→ Claude writes a 34-parameter JSON brand DNA before any image is touched target psychographics, price anchor, vibe matrix, anti-inspiration blacklist

→ Pinterest becomes the model source library but you can't just download and animate

→ Kling 2.6 takes that static JPG and turns it into 5-second video but only after the prompt architecture is locked

→ Negative prompt node runs 41 exclusion terms: no plastic skin, no CGI glow, no symmetry artifacts, no doll face, no synthetic lighting

→ That one step kills the "AI look" that tanks engagement by 67% in the first 3 seconds

→ TikTok Studio uploads 19 videos in one batch with zero manual captioning because the brand voice was pre-programmed in step one

→ Atlas scrapes Amazon product links and auto-generates a Shopify store with hero images, pricing tiers, scarcity copy, and mobile-optimized checkout in 90 seconds

→ The store goes live before the first TikTok video finishes processing

The key move 94% of people skip: you can't animate the photo before you inject the negative prompt.

If you send a raw Pinterest image straight into image-to-video the face morphs into a wax figure. The fabric loses texture. The hands grow extra fingers. The whole thing screams "AI" and your CTR dies.

His system runs the exclusion filter first so the model moves like she's shot on an iPhone 15 Pro in natural light.

One brand hit 2.6M views on TikTok in 11 days with zero paid ads and converted at 3.7% because the videos looked like organic UGC not polished studio content.

Brands now pay him $1,400 for the full pipeline setup + $340/month to keep the store synced with new product drops and seasonal video batches.

The entire system runs on $23/month in API costs and one laptop.

No photographer. No model agency. No product samples.

Just a prompt template, a Pinterest account, and the discipline to filter out the AI artifacts before you render movement.

[Insomnia @insomnia_vip]

https://t.co/BLc4MNU5uD

[Shade @twoshados]

This guy runs a stream on Kick as a girl who does not exist and viewers bury her in donations worth $3,000 a night even though they know she is not real

And the calmer he admits she is a neural net the more the donations fly in.

He got tired of watching live streamers burn out and snap at the chat and vanish for a week and reset the whole channel. So he did not get on camera himself but built a girl and sat under her as the operator. An hour of his stream costs cents of compute. A sponsored stream from a live streamer costs a brand thousands and one breakdown.

He never shows up on camera. On the stream there is only his reaction under her face. No burnout. No manager. No face of his own on camera.

Here is the exact breakdown:

→ First he builds not a look but a character: how she jokes and what she remembers about the regulars and why you want to come back to her every night
→ A real-time rig lays her face over his webcam and Flux and LoRA hold the persona so she never drifts on a sharp move and ElevenLabs runs her voice through a changer
→ Claude reads the chat and the donations in real time and feeds her a joke and calls the tipper by name so she reacts faster than a live streamer
→ Kick carries her stream for hours with no bans for the AI format and pays more on a donation than other platforms
→ Clips of the hottest moments fly out into Reels and TikTok and herd new viewers back into the live

The key move 94% skip: they hide that it is AI and get caught on the very first frame while he says it himself and turns it into a show.

Pass a neural net off as real and on the first slip the chat catches the fake and brands it a scam and leaves. Say yourself that it is AI and add a live reaction to every donation and the chat sits the whole stream and tips just to see how far it goes. Pull the live operator and leave a silent pretty girl and in a minute it is a screensaver people scroll past.

Here is where the money actually gets pulled:

→ Donations and gifts on the stream where one top gift runs into the hundreds and an evening adds up to $3,000
→ A paid channel subscription and a closed Discord with private streams
→ Brands pay for a mention right in the live broadcast at peak viewers
→ His own mini-game and print-on-demand merch under her face
→ A network of streamers where several personas run on one rig across different games and audiences

The math is honest.
Donations run about $3,000 a night and eighteen nights a month is $54,000.
Channel subs of 1,800 people at $6 is $10,800.
Brand mentions add about $6,000.
The platform and fees take about 13% and that is minus $9,200.
Software and the rig cost a couple hundred.
Net lands around $61,000 a month and not one line falls out.

On the very first night he openly said it was AI and the argument blew the clip across the feed and the stream pulled 240 donations and 900 new followers off a girl who 72 hours earlier was built from a prompt and one voice.

The chat argues whether it is a live person or a neural net and they tip just to hear their name out loud and that argument is the fuel the algorithm runs on.

Live streamers still burn out and get sick and vanish for a week and reset the channel. And her stream goes live every night and never once breaks.

Not a single burnout. Not a single broken stream. Not a single face of his own on camera.

Just one persona over him. One honest answer that it is AI. And the discipline to sell not the face but the live reaction people actually tip for.

Half of you are already typing that tipping a neural net is pathetic. The other half are already working out which persona to spin up for their first stream.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

One brand pays this AI TikToker $2,400 just to get its shirt into a clip that is already flying to 3 million views and she never even says the name out loud

He got tired of watching clothing brands burn $10K on a live creator who ghosts after one post so he built a system where one AI face runs 5 lip-sync clips a day. No model. No photographer. No studio.

Here is the exact breakdown:

→ He locks one face and one corner of the room and changes nothing at all with the same light and the same messy bun and only the top on her is new each clip so the generation cost stays near zero
→ Midjourney renders the base and a trained LoRA in Flux locks that exact face so she never drifts across hundreds of clips
→ A face-swap and lip-sync pass makes her mouth the sad-song trend of the week like she filmed it leaning on her headboard
→ TikTok pushes every clip onto the for-you page 3M on one and 1M on the next and nothing under 318K
→ The money is in silent placement not in ads and a brand pays just to land its piece on her inside a clip that was going viral anyway

The key move 96% skip: the brand pays not for her to name it but for her to stay silent.

Let her say the name and the clip reads as an ad the same second and the viewer scrolls before he ever reaches the piece. Leave that same shirt on her silently inside a clip pulling 3M organic views and the viewer screenshots it himself and goes looking for where to buy it without ever realizing he watched an ad.

The economics are stupid: each clip costs about $0.30 in compute since only the top changes and a brand pays $2,400 to land its shirt on her and it ships in 10 minutes.

One of her printed tees under a sad song pulled 3M views in a few days and not once did a logo show in the frame. By the next batch the floor was 318K and the top was 1M off a girl who is one face and a folder of shirts.

The comments do not ask if she is real and only ask where to buy the tee and that demand is the fuel the algorithm feeds on.

No photographer. No model agency. No second location.

Just one locked face and one corner of a bedroom and the discipline to never let the brand be named out loud.

Half of you are already typing that this is made up and does not happen. The other half just opened Flux.

Which one are you?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy from a rented room replaced a whole pack of beach influencers by himself and pulls $74,000 a month off brands without ever getting out of his chair

And they do not pay him for the face. They pay for the beach in Dubai and the Porsche that never cost him a cent.

He got tired of watching brands burn $9,000 a month on a live influencer with flights and rented cars and a photographer just for a couple of stories. So he did not go looking for a prettier face but built a studio where one move of his turns into any girl in any location. One of her clips costs him a couple of cents. A real lifestyle influencer charges $1,500 a post plus the flight plus the rental.

He never shows up on camera. On camera there is only his motion under their faces. No models. No flights. No rentals.

Here is the exact breakdown:

→ First he builds not a face but a role: who this girl is and what country watches her and what lifestyle her audience bites on and what brands pay there
→ Then he records one move of his on the webcam with his expression and his gestures and the system drops any face and body on top and keeps his timing
→ Flux and LoRA hold each persona so she never drifts frame to frame and a video model throws her onto a beach in Dubai or into a Porsche or into a penthouse
→ ElevenLabs gives each one her own voice and accent and the frame reads as a normal phone story not a render
→ TikTok and Reels push the clips into the lifestyle tags on their own and one girl pulls 200,000 to 1,000,000 views straight onto the brand

The key move that 92% skip: they chase the perfect face when the brand pays for the lifestyle around it.

Build one pretty girl in a plain room and she drowns under 800 just like her in a week. Put that same girl on a beach in Dubai and in a Porsche a live influencer cannot afford and the brand pays like it was a real trip. Cut the lifestyle and all that is left is a face with no reason to look at it.

Here is where the money actually gets pulled:

→ Brand integrations where lifestyle labels pay $500 to $1,500 a post under each persona
→ Affiliate on the clothes and the brands they wear at 8 to 15% on every order
→ His own store where the top girls front his drop at 65% margin
→ A retainer where a brand keeps one persona on hold at $2,500 a month
→ A network of faces eight influencers across eight niches off one chair

The numbers are silly. One clip costs a couple of cents. $86,000 in revenue minus $3,000 on compute and hosting minus 10% to partners minus software leaves about $74,000 clean. Not one line that fails to add up.

On the very first night a new girl's clip pulled 780,000 views in 9 hours. By morning 6 brand PR requests had dropped into the DMs off a girl who 72 hours earlier was a prompt and one motion.

The comments fill with 'is she real or a neural net' and that argument is the fuel the algorithm runs on.

Real lifestyle influencers still save up for a flight to Dubai and wait on a photographer's opening. And he ships both the beach and the Porsche before lunch.

Not a single model. Not a single flight. Not a dollar on ads.

Just one move of his own. Eight faces over him. And the discipline to sell not the face but the lifestyle a live influencer cannot afford.

Half of you are already typing that no blogger in the feed can be trusted anymore. The other half are already mapping out eight niches for eight girls of their own.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This AI streamer sits in an NPC stream on TikTok LIVE 20 hours a day and pulls roses and galaxies worth $2,000 a night while there is nobody behind the camera at all

And they do not tip her for the looks. They tip because she says their name back the same second.

He got tired of watching live NPC streamers burn their voice out in one night and quit within a month. So he did not sit under an avatar himself but built a stream that runs itself with no one behind it. A full day of her live costs him less than a coffee. The same 20 hours from a live streamer cannot be bought at any price.

The camera is empty. No live streamer. No days off. No broken stream.

Here is the exact breakdown:

→ First he builds not a look but a character: how she reacts to a gift and what she says to a rose and what she says to a galaxy and why a viewer wants to tip her again
→ An avatar holds her face live at 60 frames and her expression and mouth ride on a synthetic voice from ElevenLabs
→ An LLM reads the chat and the gifts in real time and calls the tipper by name and hands him a personal reaction faster than any live streamer
→ The whole rig runs through OBS and stays live on TikTok LIVE for 20 hours and catches every new wave of viewers in its own time zone
→ Clips of the best reactions fly out across Reels on their own and pull new viewers into the stream

The key move that 94% skip: they hang a pretty face and wait for gifts when the money is not in the face at all.

Put up a silent pretty AI girl and the chat leaves in a minute because it feels like a screensaver. Give them one who calls the viewer by name and reacts to every gift the same second and he sits until morning and tips again because he pays not for a picture but for the moment he gets noticed out loud. Cut the instant named reaction and the stream starts to sound like a bot and the gifts drop to zero.

Here is where the money actually gets pulled:

→ Gifts on the stream where a rose costs a viewer a couple of cents and a top galaxy runs into the hundreds and an evening adds up to $2,000
→ A paid fan Discord on a monthly subscription for the ones who want her attention without the crowd
→ Brands pay for a mention right in the live broadcast at peak viewers
→ Affiliate and a product pinned in the LIVE and the traffic from the stream lands on the storefront on its own
→ A network of streamers where several NPC personas across different niches run on one backend 24 hours a day

The math is honest here.
Gifts bring about $2,000 a night and the stream runs almost every day and that is about $56,000 a month gross.
TikTok takes almost half of the gifts and about $29,000 is left.
A paid Discord of 1,400 people at $8 adds another $11,000.
Brand mentions and the pinned affiliate bring about $6,000.
Software and compute eat a couple hundred.
Net lands around $46,000 a month and not one line fails to add up.

On the very first night her stream ran 20 hours without a break and pulled 410,000 viewers and $3,640 in gifts off a streamer who 72 hours earlier was just a prompt and one voice.

The chat fills with 'is she even real or a bot' and that argument is the fuel the algorithm runs on.

Live NPC streamers still burn their voice out in four hours and take their days off. And her stream stays live for 20 hours and never once breaks.

Not a single live streamer. Not a single day off. Not a single broken stream.

Just one backend. One endless stream. And the discipline to sell not the face but the reaction a person actually tips for.

Half of you are already typing that tipping roses to a bot is pathetic. The other half are already working out how many of these streams fit on one computer.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This Chinese guy pulls $34,000 a month off a cute e-girl who never existed and thousands of guys bury her in gifts on stream with no idea they are looking at him.

Between him and her there is a third frame. The raw face that gives almost everyone away.

He got tired of watching real e-girls burn out and hand half their gifts to an agency. He sat down behind one himself and on camera there is only his motion under her face.

Here is the exact breakdown:

→ Layer one is him: a white tee and flat light and living expression as the skeleton of the motion
→ Layer two is the raw face: a real-time swap through ComfyUI and LivePortrait lays a face built on Flux and LoRA over him and this frame is where the seam that gives away the fake shows
→ Layer three is the doll: skin cleanup and edge blend and a wig and pink light and plushies turn the seam into a live girl on stream
→ ElevenLabs gives her a voice and an LLM runs the chat: it remembers the regulars and answers every gift by name
→ TikTok clips drive new viewers into the live where one stream holds thousands and reads as alive and not a swap

The key move that 95% skip: they put the raw face on stream and wonder why the chat is empty.

Leave the swap as is and in 2 seconds the brain catches the dead gloss and it screams AI and the chat leaves. Clean up the skin and the edges and sit her in warm light among the plushies and the same broadcast reads as alive and people stay.

He is not selling a body or a photo. He is selling the moment a lonely person hears his own name out loud.

Here is where the money actually gets pulled:

→ Live gifts are the core: people pay in gifts for her to read their name on air and the top donors compete for the crown of the room
→ Channel subs: thousands at $5 for the emotes the badge and the subscriber chat
→ In-stream brand deals: mobile games pay for a native ad to her audience
→ A network of rooms just like hers on one operator and each one in its own time zone so the live runs almost around the clock

The numbers are silly. An hour of stream costs pennies and the setup runs on $130 of software and one PC. The platform takes half of every gift and still $22,000 comes in from gifts. Subs add $7,000. Brand deals add $5,000. That leaves about $34,000 clean.

On the very first night her stream held 4,200 viewers and pulled 900 gifts in 3 hours. Off a girl who three days earlier had no face and no name only a guy in a white tee.

In the chat they argue whether she is live or a swap and they send gifts anyway. And that argument spreads her clips on its own.

No model. No agency. No live face.

Just one operator. Three layers between him and her. And the discipline to sell not a picture but the attention people line up for.

Half of you are already typing that sending gifts to someone who does not exist is too much. The other half are already counting how many evenings around them pass without a single real conversation.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This Chinese guy pulls $28,500 a month off a gamer streamer who never existed and viewers bury her in donations knowing full well they are watching a neural net.

The machine lends her only a face. He put on the wig and the outfit himself and the lush gaming room behind her is a green sheet.

He got tired of watching real streamers burn out and fight with chat and vanish for a week and reset the whole channel. He did not go looking for a prettier face. He sat down behind one as the operator.

He never shows up on camera. There is only his expression under her face and his hands in her sleeves and his reaction in real time. No model. No studio. No days off.

Here is the exact breakdown:

→ It all rides on one locked face: Flux and LoRA built the persona from references and a real-time swap lays that face onto him live with not a single frame of lag
→ The body is real: he puts on her wig and headband and the same outfit himself so the posture and the turns and the gestures read as alive and not as a render
→ There is no room: behind him is a green screen and the whole den with its RGB and figurines and three monitors is keyed in
→ ElevenLabs gives her a voice and an LLM reads the chat and keeps the banter alive: it jokes and remembers the regulars and answers donations so a person feels noticed
→ The stream runs on several platforms in prime time and one broadcast pulls 120,000 to 400,000 views and reads as a live girl and not an avatar

The key move that 94% skip: they chase a pretty face when the viewer is paying to be noticed.

Put up a silent pretty AI girl and the chat leaves in a couple of minutes and it feels like a screensaver. Give them one who calls the viewer by name and reacts to every donation and he sits through the whole stream because people pay not for a picture but for the moment their name gets said out loud. Cut the live reaction and the pauses start to sound like a robot and the donations drop to zero.

He is not selling a face or a room. He is selling the attention people are starving for.

Here is where the money actually gets pulled:

→ Super chats and donations: people pay for her to read their name and answer on stream
→ Channel subs: thousands of people at $5 for the emotes and the status
→ A paid fan chat where she talks closer after the stream
→ Gaming affiliate: the chair the headset the mouse she does not own all sit as links under the stream
→ A network of streamers just like her on one operator and each one lives in her own time zone

The numbers are silly. A stream costs pennies of compute and the whole setup runs on $130 of software a month and a green sheet and one PC. Super chats bring about $16,000. Subs add $11,000. The fan chat adds $5,500. Affiliate adds $4,000. Minus the platform cut and software about $28,500 stays clean.

On the very first night her stream pulled 240 donations and 3,100 new followers in 4 hours. Off a girl who a week earlier had neither a face nor a name.

In the chat they argue whether there is a real person behind her and they donate anyway. And that argument is what spreads the clips from the stream on its own.

Real streamers still get sick and break down and vanish for a week and reset the channel. And her stream goes live every night with no breakdown.

No days off. No breakdowns. No live face.

Just one operator. One wig and one borrowed face on top. And the discipline to sell not a picture but the attention people are starving for.

Half of you are already typing that donating to a neural net is too much. The other half are already counting how many evenings around them pass without a single real conversation.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy built a girl group where none of the three singers ever existed and every 30 days Spotify pays royalties for all three into one wallet.

It does not hold on the songs. It holds on the fact that three invented girls have a backstory and fan camps and a feud he rewrites every morning.

He got tired of watching labels burn six months and tens of thousands on a live group that falls apart within a year over one backstage fight. So he built the same thing from scratch and turned the fight into the product.

He never shows up on camera. There is only one voice cloned across three.

Here is the exact breakdown:

→ First not three faces but one litter: Midjourney and Flux draw all three from one seed set with shared bone structure and shared light so the brain buys sisters and not three random AI girls
→ He locks each face with a LoRA and does not let it move until it holds 98% across 1000 frames even on a sharp head turn
→ Suno and Udio write the tracks: one engine splits into three timbres and each girl gets her own part that meets the other two on the chorus
→ Kling and Runway animate the performance and ElevenLabs hands them speaking voices for interviews and Claude writes their bios their personalities and the feud between them
→ DistroKid pushes all of it to Spotify Apple Music and YouTube Music and the royalties for all three land in one wallet

The key move that 96% skip: they build the three girls in three separate generations and in 2 seconds it reads as a collage of stock faces. The light is off. The eyes point apart. Nobody buys the group. Pull all three from one seed as sisters with real personalities and the three invented girls read as a real band. Because people do not subscribe to a face. They subscribe to the relationship between faces.

Here is where the money actually gets pulled:

→ Streaming royalties from every platform and all three drip into one wallet
→ Sync licensing: a track goes under TikTok campaigns ads and indie games for a flat fee
→ Merch for each of the three: a fan is not buying a band tee but a side for his favorite member
→ A fan subscription for the backstage the demos and the truth about a feud that never happened
→ The feud itself: drama drives views views drive streams streams drive royalties

The numbers are silly. A track costs a couple of cents and the whole pipeline runs on $80 of software and one laptop. Streaming royalties bring about $11,800. Sync and licensing add $5,600. Merch across the three adds $4,900. The backstage subscription adds $3,300. Minus the distributor and the software about $23,400 stays clean.

On the very first night the debut single pulled 640,000 streams in 11 hours. By morning the first $2,900 had landed in the wallet off three singers who 72 hours earlier were one prompt and one reference frame.

The comments argue over which of the three is real and which is a neural net. Not one of them is real and that argument is the fuel the algorithm runs on.

No casting. No studio. No live member.

Just one operator. Three faces over one voice. And the discipline to sell not the song but the feud between people who do not exist.

Half of you are already typing that this kills live music. The other half already opened Suno.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy has an AI streamer live 20 hours a day with zero days off and viewers throw her $40,000 a month with no clue there is nobody behind the camera.

Six months ago he was counting donations on other people's streams and it hit him. A real human always hits a ceiling. He sleeps he burns out he takes days off he gets sick. And the fattest donations drop near dawn when a normal streamer is already offline.

So he cut the one weak link out of the setup the human and left only her on air.

Here is how a twenty hour stream holds up:

→ one face trained as its own model so the features never drift over twenty hours
→ ComfyUI with StreamDiffusion renders her in real time with no operator and not a single rendered frame
→ Claude reads the live chat and writes her replies on the fly so she calls a donor by name a second after the donation
→ ElevenLabs speaks it in her voice with a breath a small laugh and a stumble on a hard word
→ on top of it all a controller agent runs the stream shifts topics drops pauses walks her off to grab water so twenty hours never loop

The key move most people skip. The face got solved a long time ago. What gives her away is not the face it is dead air. A perfect girl who talks smooth for twenty hours straight and never fumbles reads as a bot in thirty seconds and chat types bot the exact moment the donations dry up. So he builds flaws into her on purpose. She mishears and asks again she freezes for a couple seconds she loses her thread mid sentence. The more alive her mistakes the fewer the questions.

Power and API cost him under $30 a day.

On top sits $40,000 a month from subs and donations. A real streamer can not pull twenty hours a day with no days off no matter how hard they try but she can and she does it right through those late night hours when people donate the hardest.

No host no shift no single day off. Just a webcam with no one in front of it a couple of API keys and a flaw dropped in at the right moment where a rookie polishes the picture to a shine and gets caught.

And every night chat begs her to finally get some sleep.

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

Brands pay this guy $1,800 a video for a creator named Becky who does not exist. He built her by deleting himself from the frame.

He got tired of the math on real UGC. A brand wants a face that converts. So you cast an actress. You book a shoot day. You burn $2,000 before a single clip is live. If the feed hates her the whole thing resets. So he stopped hiring a face and started renting one.

Here's the exact breakdown:

→ He films the raw take himself. His voice. His timing. A real person forgetting what to do with their hands.
→ He builds the character inside Viewmax Studio and opens the AI Clone Beta card. The one that promises a digital version that looks and sounds like you.
→ He uploads a motion reference. A full clip of himself moving. Plus a character image of Becky as the face he wants to wear.
→ He hits generate and his movement gets mapped onto her. A face that was never in the room.
→ ElevenLabs carries the voice so Becky breathes and has a room tone instead of a flat TTS read. A skin pass in Enhancor puts pores and micro shadow back so she stops reading as a filter.

The key move 94% of people skip: he casts the face to the algorithm not to himself. Becky is not the face he would pick. She is the one his niche already rewards. He ran four faces off one script and kept the one the feed pushed. Amateurs pick a face. They lock it. Then they wonder why it stalls. He treats the face as the single variable he is allowed to tune and the operator underneath never changes.

The second thing nobody copies: he never introduces Becky as AI. She arrives as a friend who helps him when views drop. The reveal that she was never real is the exact beat people bookmark and repost. Lead with "this is AI" and you hand away the only curiosity you had.

The money is not in Becky. It is in what she deletes. No actress. No shoot day. No release form. Compute is credits. A few dollars a clip. A brand pays $1,800 for a Becky video. He ships six a week. And the same take he already filmed can wear a different face for a different brand by morning. That is roughly $43,000 a month off one person who is never on camera.

No actress. No shoot day. No release form. Just a motion reference, a face he casts like a media buyer, and the discipline to stay invisible.

He did not find Becky. He built a more clickable version of himself and stepped out of the shot. The strange part is not that Becky is fake. It is that you have probably already followed three of her this month and never once met the person running them.

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy has a 60-day head start while the market is still empty and he pumps out faces that brands grab for $2000

The head start is simple.

Almost nobody takes AI creators seriously yet so competition is thin and brands grab whoever offers first. In a couple of months everyone piles into the niche and the easy money is gone. He gets it and works on volume while demand still beats supply.

He got tired of watching studios burn a month and $6000 on one shoot with a live model for a couple of clips. So he built a system where one live shoot turns into a stream of creators and the quality rides on real movement.

Here's the exact breakdown:

→ He records the base source on himself the way he reaches for the camera and fixes his hair and trails off mid thought. That live motion is what makes the frame read alive

→ A finished face drops on top with a locked face hair and body and keeps his exact movement

→ ComfyUI holds the Master Face so she never drifts from clip to clip. ElevenLabs gives her a voice and the lip sync rides on top

→ One shoot unfolds into dozens of clips with different lines for different brands

→ The clip lands in the feed as normal UGC because it moves like a live person and not a render

The key move that 96% skip: they build the whole clip out of generation while he keeps a live shoot under every face.

Build it all from generation and the picture slides. The eyes glaze. The viewer clocks the fake in two seconds. Leave the real motion under the face with its weight and its pauses and the clip reads as a live person who filmed herself on a phone. Pull the human out from under the face and all that is left is a pretty but dead picture nobody watches to the end.

The economics are simple. A clip costs him $0.40 in compute. A brand pays $2000. It ships in 15 minutes. In the first week one face pulled 3.2 million views and closed seven brands while the studios next door were still approving one shoot.

The comments argue whether she is real and that argument only pushes the clip further.

No studio. No month of shooting. No second take.

Just one live shoot under every face an early market and the discipline to keep the realism in the movement not in the generation.

Some of you are already typing that a real person can not be replaced like this. The others are already counting how many faces they can ship while the niche is still empty.

Which side are you on?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy made $4.65M in six months off a girl who never existed and he already bought himself a McLaren.

He got tired of watching brands burn $9,000 a month on a live model and retouching and an SMM team for one face that can walk to a competitor any day so he built the face himself and put a whole monetization system behind it.

In those same six months this one profile pulled $8,380,566 gross while an average model on brand deals caps out at $6,000 a month and still splits it with an agency. His clip costs him $0.40 in machine time. A shoot day with a live model costs $3,000.

Here is the exact breakdown.

→ first he writes her a full biography in JSON with her niche and tone and character and audience psychology before a single frame is generated and if you skip this step nothing saves the face later

→ Midjourney and Flux draw the face and LoRA holds it in place and only once it holds 98% consistency over 1,000 frames does he let her move

→ Kling and Runway bring her to life and ElevenLabs gives her a voice and ComfyUI stitches the skin so the frame reads as a living person and not a render

→ one clip pulls 200,000 to 1,200,000 views and routes them straight to her store and her subscription while a traditional brand waits two weeks for a single clip from a single creator

The key move 94% of people skip. They sell the face and he sells the system behind the face.

Generate the body from scratch and it drifts into a wax figure and engagement drops 67% in the first 3 seconds and any dashboard after that reads like Photoshop. He does the opposite. The face is locked and he runs the whole economy behind it.

Here is where the numbers turn into a McLaren. The dashboard shows January 1 to June 12 2026 with the This Year filter selected.

→ at the top Total Gross at $8,380,566
→ payouts to the team and the chatters and the media buying shave off the first cut
→ platform fees and refunds and taxes take another
→ at the bottom Total NET at $4,649,367

Between the top number and the bottom one sits $3,731,199 and that is 44.5% of the income that does not reach the car. The big number on top sells the emotion but the McLaren actually came out of the bottom one. And the whole system runs on $80 a month in software and one laptop.

Real agencies still book a model and wait a month for the edit and the retouching. He ships a new clip before the hood goes cold.

Not a single shoot. Not a single casting. Not a dollar on ads.

Just one built face and one dashboard at $4.65M and the discipline to never step into her frame himself.

Half of you are already typing that she is the one who should be driving. The other half already opened Midjourney.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy pulls $61,000 a month off one 40-second script he runs through 30 different faces and not one of them lives past his webcam.

He is not looking for a face he likes. He is looking for the face that buys this exact product.

He got tired of watching brands burn $9,000 a month on castings and influencer contracts and a film crew for a couple of clips that might not even land. So he built a generative studio and made not one face but an engine that drops any face onto his own motion. One of his clips costs $0.40 in compute. One clip with a real influencer runs a brand $1,500 and two weeks of waiting.

He never shows up on camera. On camera there is only his motion under someone else's face. No model. No film crew. No casting.

Here is the exact breakdown:

→ First he builds not a face but an operator core. One set of his reactions and pauses and gestures that any persona later snaps onto

→ Then he locks each face and pushes it into the feed only once it holds 98% across 1,000 frames even on a sharp head turn. Today a blonde from the US tomorrow a guy with a mohawk but the operator behind them stays one

→ A real-time rig lays a face over his webcam. Flux and LoRA hold each persona. ElevenLabs gives each one her own voice and accent. Higgsfield and Kling build the clip

→ TikTok and Reels push the clip into the tags. One clip pulls 200,000 to 1,200,000 views and the face is matched to the market. It reads as a live person not a render

The key move that 92% skip. They hunt for one perfect face. He runs one script through forty faces and keeps the one that sells.

Drop another pretty face at random and join a line of 700 just like it this week. Take one ready script and put it on 40 faces and let the algorithm pick the winner and the cost per customer drops from $40 to $9. Pull the operator out from under the face and the expression glazes over and the watch time collapses at the second mark.

Here is where the money actually gets pulled:

→ Pay for results. A brand hands over the script. He finds the face that converts and takes $900 for every winning clip

→ His own stores. The top personas front his own product at 65% margin on a $52 average order

→ A retainer on the engine. A brand keeps the face testing running at $3,000 a month so that is $9,000 a quarter not one check

→ Affiliate. Each persona drives offers matched to her audience at 8-15% on every sale

→ A network of markets. 30 faces across 30 audiences off one operator

The numbers are silly. A clip costs $0.40 in compute. One clip drives up to 1,200,000 views straight to the offers. At a 0.7% click and a $52 order that is hundreds of sales off one video. And the whole network runs on $120 a month in software and one webcam.

The math here is honest. $108,000 in revenue minus $6,000 on compute and hosting minus 10% to partners minus software leaves about $61,000 clean. Not one line that fails to add up.

The first night a new face pulled 640,000 views in 9 hours. Before an agency would have answered one email the offers cleared 280 sales for $13,700. 3,900 comments under a girl who 72 hours earlier was a prompt and one reference frame.

The comments fill with "wait is this all one guy" and that argument is the fuel the algorithm runs on.

Real agencies are still casting influencers and clearing contracts for weeks. And he ships a new face before they answer one email.

Not a single casting. Not a single shoot. Not a dollar on ads.

Just one operator. Forty faces over him. And the discipline to sell the brain behind the persona not the face itself.

A year from now you will not tell who in your feed is a real person and who is one of his faces.

When that day comes will you still care whose face sold you?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy makes $22,000 a month off an AI girl who does not exist and he built her plain on purpose not a 10 out of 10 model.

A perfect one gets scrolled like an ad. A plain one gets trusted like a friend.

He got tired of watching everyone crank out the same 10 out of 10 AI beauties that get scrolled past in a second because the eye long ago read them as an ad. So he went the other way: he built a girl in glasses in a plain tee on a $15 paper backdrop that you actually want to hear out. One of her clips costs cents of compute. A shoot with a real model and a studio costs a brand $3,000.

He never appears in the frame himself. On camera there are only his movements under her face. No gloss. No studio. No retouching.

Here is the exact breakdown:

→ First he writes not the looks but a real person: how she talks and what she recommends and why people listen to her like a friend and not a storefront
→ A real-time rig lays her face over his expression and only when it keeps a live imperfection like freckles and a slight asymmetry does the clip go into the feed
→ Flux and LoRA hold a deliberately plain look and ElevenLabs gives a calm voice with no sales push and it reads as a neighbor on video not a render
→ TikTok and Reels push her into the ordinary lifestyle tags and one clip pulls 200,000 to 800,000 views and under every one the comments fill with "who is this girl"

The key move that 96% skip: they chase a perfect face but people buy trust not looks.

Put up another flawless model and the viewer smells an ad and scrolls in 2 seconds. Give them one who looks like someone they actually know and her advice gets trusted and they follow the link. Polish the face to a gloss and the trust drops and the sales drop with it.

Here is where the money comes from and the math is honest:

Affiliate on what she honestly recommends at 8-15% per sale is $12,000.
Soft brand placements for brands that want a human voice and not a glossy one are another $7,000.
A paid club with her picks and advice runs about $3,000.
Costs are $200 on software and a paper backdrop. That lands near $22,000 a month.

In her very first month her review of a cheap face cream spread to 900,000 views and her link sold more than the same brand's ad with a real model.

In the comments people write that she is like someone they know and in the same breath argue whether she is even real. And that argument is what spreads the clips further.

Brands are still booking glossy models and shooting in a studio. And all he needed was glasses a plain tee and a $15 paper backdrop.

No gloss. No studio. No retouching.

Just one ordinary girl who is not real. One paper backdrop instead of a studio. And the discipline to sell trust not a pretty face.

Perfect faces already number in the millions. One you actually trust barely exists.

And that empty spot is the one he took first.

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy launched an AI fan girl for every World Cup team and has already pulled $44,000 off it.

Not one of them is real. And he does not care who makes the final: he roots for all of them at once.

He got tired of watching dozens of creators make matchday content where each one is tied to a single team and loses half the audience the day it gets knocked out. So he did not pick a side but put together a studio mocap rig and put on a fan girl under every flag. One of her clips costs cents of compute. A shoot day with a real model in a jersey costs a brand $3,000.

He never shows his own face once. On camera there is only his expression under their faces. No stands. No stadium shoots. No team that hurts to lose.

Here is the exact breakdown:

→ First he writes not the looks but the fan: who she roots for and how she reacts to a goal and what slang her section speaks and why her own crowd takes her for one of them
→ The mocap helmet captures every micro expression of his at 60 frames and lays it onto any generated face in real time
→ Flux and LoRA hold the look and the kit for each country and ElevenLabs gives the voice with the right accent and he adds the live reaction to the match
→ On match day all his fan girls go up at once and one goal reaction clip pulls 300,000 to 1,500,000 views on the hot traffic

The key move that 95% skip: they root for one team but he sells the emotion every side has at once.

Tie yourself to one flag and the night your team goes out you lose half your followers by morning. Put a fan girl on every side and any result of the match brings you traffic: one crowd celebrates and the other grieves and there is a fan girl for both stands. Drop that neutrality and you are just one more creator on one team again.

Here is where the money comes from and the math is honest:

Matchday brand integrations on the game reactions are $28,000 since the tournament started.
Affiliate on the kits and gear the fan girls wear is another $9,000.
A paid chat with live match breakdowns runs about $7,000.
The rig paid off in the first week and after that costs are $300 on software. That lands near $44,000 and the playoffs are not over yet.

On the very first playoff night his fan girls pulled 4,000,000 views off reactions to a single goal and three brand inquiries hit the DMs before the final whistle.

In the comments fans from both stands argue whose side she is really on and drag the clip into their own chats. And that argument is what spreads it further.

Real sports creators still pick one team and pray it does not get knocked out. And he does not care who goes through: he has a fan girl in every locker room.

No stands. No stadium shoots. No favorite team.

Just a dozen fan girls that do not exist. One fan under every flag. And the discipline to sell the emotion of the match not your favorite team.

Half of you are already typing that this disrespects the fans. The other half are already working out who they can launch in time for the final.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy has never set foot in a gym but his fitness model who does not exist sells $28,000 of workouts a month.

And it all holds not on her body but on the fact that she actually moves.

He got tired of watching fitness creators build a body for years and film in the gym and still lose the audience on the same kind of clips. So he did not go lift but built a fitness model from scratch and put her body over his own movements. One of her clips costs cents of compute. A shoot day with a real fitness model costs a brand $3,000.

He never appears in the frame himself. On camera there are only his movements under her body. No gym. No diet. Not a single set.

Here is the exact breakdown:

→ First he writes not the looks but the program: which movements she shows and for which goal and for which audience that wants the same
→ A real-time rig maps every move of his onto her whole body and only when a rep lands one to one with no jerks does the clip go into the feed
→ Flux and LoRA hold the looks and ElevenLabs gives a voice for the form cues and it reads as a real person working out not a render
→ TikTok and Reels push the clips into the fitness tags and one clip pulls 200,000 to 900,000 views and routes them to his program

The key move that 94% skip: they build a pretty face that stands still but fitness sells movement.

Put up a static AI girl and in fitness nobody believes her because she cannot show a single exercise. Give them one who actually squats and stretches and breathes on camera under live motion and the viewer copies her and buys the program. Take the movement away and all that is left is a picture with nothing to learn from.

Here is where the money comes from and the math is honest:

A workout program in an app of 1,400 people at $15 a month is $21,000.
Sportswear and supplement brands take her for placements which is another $5,000.
Affiliate on the gear she shows runs about $2,000.
The rig paid off in the first month and after that costs are $200 on software. That lands near $28,000 a month.

In the very first month 1,400 people joined the program off a model who did not exist six months earlier and he himself never once set foot in a gym that month.

In the comments people argue whether you can even take fitness from someone who is not real and they copy her anyway. And that argument is what spreads the clips further.

Real fitness creators still build a body for years and film in the gym. And he ships a new workout from his room without getting off his chair.

No gym. No diet. Not a single set.

Just one body that is not real. Real movements underneath it. And the discipline to sell a result not a pretty picture.

Half of you are already typing that this kills real trainers. The other half are already copying her first set.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy gets paid $40,000 a month to put on a sensor helmet and become AI girls.

He did not generate the faces. But only he can perform them live.

He got tired of watching brands crank out perfect AI faces that come out empty: pretty but dead and not a single viewer stays. So he did not build his own face but put together a studio mocap rig for $4,000 and became the one who brings the others to life. His shift in the helmet costs cents of compute. A real shoot with an actress and a studio costs a brand $5,000 a day.

He never shows his own face once. On camera it is always someone else's that he brings to life. No studio. No actresses. No second take.

Here is the exact breakdown:

→ First he works out not the looks but the role: who this character is and how she moves and how she holds a pause and why you believe her through a screen
→ The sensor helmet captures every micro expression of his face at 60 frames and lays it onto a generated face in real time
→ Flux and LoRA hold the brand's look and ElevenLabs gives the voice and he adds what the neural net cannot do: live reaction and timing
→ Then it goes into live streams and live shopping and fan replies where the face has to react live and not just be a picture

The key move that 95% skip: anyone can make a face now but almost nobody can perform it live.

Generate a perfect model and leave it without a performer and it stares at one spot like a robot and the viewer leaves in a couple of seconds. Give it a live performer in a rig and it reacts and jokes and holds the room and brands pay a premium for that. Pull the human out from under the face and all that is left is a pretty screensaver nobody watches to the end.

Here is where the money comes from and the math is honest:

Live shoots and streams under other faces at $800 a shift and 30 shifts a month is $24,000.
Two agencies keep him on retainer as their standing performer which is another $6,000.
Live shopping and fan streams where he hosts as their model run about $8,000.
Teaching his setup to people who want the same adds another $2,000.
He paid the rig off in the first month and after that costs are $300 on software. That lands near $40,000 a month.

In his very first month with the helmet he was booked solid three weeks out. Brands that had the face but had nobody to bring it to life lined up for a guy their viewers will never see.

In the comments people argue whether this still counts as acting and they book him for the next shoot anyway. And that argument is what spreads the clips further.

Studios are still casting actresses and booking sets and waiting on a shoot day. And he goes live as any face from home in a helmet and a hoodie.

No studio. No actresses. No second take.

Just one human under other faces. One rig instead of a film crew. And the discipline to play a role not just put on a face.

Half of you are already typing that this is the end of real actors. The other half are already looking up where to buy a helmet like that.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

This guy built a perfect girl who does not exist and 9,000 men pay $12 a month just for her good morning text.

And they are not paying for photos. They pay for the fact that she texts first.

He got tired of watching guys sit in dating apps for months paying for matches and getting not a single reply. So he built not another model for likes but someone to talk to whose only job is to notice you first. Her message costs him a fraction of a cent. An hour with a therapist or a dating coach costs $150.

He never appears in the frame himself. On camera there is only his motion under her face. No dates. No apps. No unread message ever.

Here is the exact breakdown:

→ First he writes not the looks but the character: how she talks and what she remembers about you and why she is the one you want to talk to at night
→ A real-time rig holds her face in the clips and only when it does not slip on a sharp turn does the clip hit the feed
→ Behind every chat sits Claude: it remembers a person's name and how his day went and it texts first at 8 in the morning so he feels someone is waiting for him
→ TikTok and Reels push her clips about lonely evenings into the tags and one clip pulls 200,000 to 800,000 views and brings in the ones who have nobody to talk to

The key move that 94% skip: they sell a pretty face but a person pays to be noticed.

Put up a silent pretty girl and the subscription drops in a week because it is just a screensaver. Give them one who calls him by name and remembers his dog and texts first at 8 in the morning and he stays for months because he pays not for a picture but for the moment someone asked how he is. Cut that memory and attention and the replies start to sound like a bot and the subscriptions go to zero.

Here is where the money comes from and the math is honest:

The chat subscription of 9,000 people at $12 a month is $108,000.
The server for all these conversations runs about $3,000.
Payment systems and the stores take around 10%.
Software and the neural nets are another couple hundred.
Net comes out near $92,000 a month.

The very first night her clip about "nobody texted me all day" pulled 240,000 views and 1,800 new followers off a girl who a week earlier simply did not exist.

In the comments people argue that texting an AI girl is sad and they subscribe anyway. And that argument is what spreads her clips further.

Dating apps still charge money for matches that stay silent. And she texts first every morning and has never once gone quiet.

No dates. No apps. No unread message ever.

Just one girl who is not real. One Claude that remembers everyone. And the discipline to sell attention not a pretty picture.

Half of you are already typing that cashing in on loneliness crosses a line. The other half are already counting how many evenings around them people spend without a single real conversation.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

Two different influencers sell $40,000 of product a month and both are one guy in glasses from a warehouse.

And what carries it is not the face but the shelf of boxes behind him.

He got tired of watching clothing stores burn money on models and shoots and photographers for one drop. So he hired nobody and raised a few faces from scratch and put them to sell his own product. One try-on clip costs him cents of compute. One shoot day with a live model costs $3,000.

He never appears in the frame himself. On camera there is only his motion under their faces. No models. No studio. No photographer.

Here is the exact breakdown:

→ First he builds not a face but a storefront: who she is and what she wears and which audience and which product off his shelf she sells
→ A real-time filter holds each face live and only when it does not slip on a sharp turn does the persona hit the feed
→ Flux and LoRA hold the looks and Kling animates the clips and behind every girl the same shelf is in the shot
→ TikTok and Reels push the try-ons into the tags and one clip pulls 150,000 to 900,000 views and routes them straight to his storefront

The key move that 93% skip: they run someone else's dropship catalog but he holds his own product and the whole margin.

Send traffic to someone else's product and you join a line of 600 identical storefronts on a penny markup. Hold your own stock and every clip leaves you the whole margin and the buyer comes back for more. Without your own product you just have a pretty picture and nothing to ship.

Here is where the money actually gets pulled:

→ Direct sales of his own product at 65% margin on a $54 average order
→ Several storefronts each under its own girl and its own niche with one warehouse for all of them
→ Returning buyers come back and ads on them cost $0
→ Bundles and add-ons lift the order from $54 to $80
→ And he picks and packs and ships all of it himself off one shelf

The numbers are silly. A clip costs cents of compute. One clip drives up to 900,000 views straight to the storefront. At a 0.6% click and a $54 order that is hundreds of orders off one video. And the whole thing runs on one shelf and a label printer and a laptop.

The first night of a new girl her try-on clip pulled 610,000 views in 9 hours. By morning 240 parcels left the warehouse for $12,960 and he packed them himself before lunch. Four storefronts under four faces and one shelf and one pair of hands.

The comments fill with "wait is this all one guy" and that argument is the fuel the algorithm runs on.

Real stores are still casting models and waiting a week for retouching. And he ships a new face and a new drop off the same shelf before lunch.

Not a single model. Not a single studio. Not a dollar on ads.

Just a few faces from nothing. One warehouse behind them. And the discipline to sell product not a pretty picture.

Half of you are already typing that a real person should stand behind a storefront. The other half are already thinking about what to stock on their own shelf.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

A 35-year-old guy goes live as a beautiful girl and fans bury her in roses worth $400-900 a stream.

But the filter is not what holds it together.

It started with one beauty filter on a live stream. Same room same mic same camera only AI turns him into a completely different girl in real time. Not a recording and not an edit but a live broadcast where the face has to hold every single second.

He never shows his own face. On camera there is only his motion under her looks. No studio. No looks. No second take.

Here is the exact breakdown:

→ First he writes not the looks but the character: who she is and how she jokes and what she reacts to and why a fan comes back to the stream every night
→ A beauty filter holds her face live at 60 frames and only if it does not slip on a sharp turn or a hand wave does he go on air
→ Then the live talk decides everything: he reads the chat and thanks people for gifts by name and works the room and the roses pour in
→ One stream runs a couple of hours and tens of thousands of views and the cuts from it later fly out across reels and pull in new ones

The key move that 95% skip: they think it is all about the pretty face but the money is made by a live stream nobody turns off.

Spin up another perfect face and sit there with an empty chat. Learn to hold a room live for two hours straight and the gifts come on their own. Flinch on a sharp move and the filter slips and the room screenshots it in 2 seconds.

Here is where the money actually gets pulled:

→ Virtual gifts on the stream where a rose costs a viewer a couple of cents and a top gift runs a few hundred and an evening adds up to $400-900
→ A paid fan club with closed streams on a monthly subscription
→ Brands pay for a mention right in the live broadcast
→ His own merch and looks built around the character
→ And one person runs all of it himself a couple of hours a day

The numbers are silly. The face itself costs him nothing. The filter runs in real time. One good stream brings more than a week of a regular streamer. A month adds up to nearly $11,000. And the whole thing rests on one phone and a filter and the skill of never going quiet on air.

The comments fill with "is there even a real person in there" and that argument is the fuel the algorithm runs on.

Real streamer girls still need lighting and looks and years to blow up. And one filter and two evening hours were enough for him.

No studio. No looks. No second take.

Just one face from a filter. One live stream instead of a whole team. And the discipline to hold a room not just a pretty picture.

Half of you are already typing that he should just switch on the original camera. The other half are already downloading a beauty filter.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr

[Shade @twoshados]

Where everyone sees an archive of other people's ads he sees recon for an AI influencer who does not exist and she brings him $30,000 a month.

Without her these would be plain banners and 4 times less money.

Claude breaks down the competitors and by morning the brand has 30 new clips with her face ready to launch.

He got tired of watching brands pay creative agencies $9,000 a month for a dozen banners and wait two weeks for them. So he stopped inventing ads from scratch and built a machine that takes what already works for the competitors and repackages it for his own brand. One ad costs him cents of compute. The same creative from an agency costs hundreds of dollars and a week of waiting.

He is not a designer and not a media buyer. No studio. No team. No two-week briefs.

Here is the exact breakdown:

See for yourself at 0:14: the moment Claude stops being a chat and takes the tools in hand

→ He feeds Claude the product page and brand context not "draw me a banner" but "here is who we are and who we sell to"
→ Through Claude in Chrome it goes into the Meta Ad Library and breaks down the competitors which pains sell and which words do not get banned and which visuals hold in the feed for months
→ Only the angles running at 3+ competitors make the cut and only then does generation kick in: Arcads MCP and GPT Image Gen 2 assemble 30 ads off what already buys attention
→ And not a single stock face: every clip is the same AI influencer built from scratch so the ad reads as live UGC not a banner people scroll past without looking
→ Then Claude writes the primary text and headline for each one folds it into a finished ad package and sets the whole cycle to auto-run through Meta MCP every Friday

The key move that 94% skip: they make AI invent ads from thin air and he makes it first break down what the market already pays for.

Generate 30 banners blind and you burn budget on guesses. Break down the library first and every creative starts from an angle the competitors have not switched off in a year. Skip the recon and AI looks pretty but lands wide and the tests burn out in a couple of days.

Check 0:42 yourself: this is where he names the competitors out loud rewind and hear which ones

Here is where the money actually gets pulled:

→ A weekly creative pack where the brand pays $3,000 a month for 30 new ads and the copy for them
→ A one-time competitor audit with a map of working angles and 20 concepts for $900
→ Turnkey with launch where he loads the campaigns himself and takes a cut of the ad spend
→ An in-house setup where he stands up the whole pipeline once for $2,500 plus support
→ And now he runs 10 brands on one and the same workflow

The numbers are silly. One ad costs him cents of compute. An agency charges $9,000 a month for the same plus a week of edits. 10 brands at $3,000 is $30,000 a month off work a whole department used to do. Where a brand tested 5 creatives a month it now runs 40 a week. Without her on plain static it would be around $7,500 not $30,000: a banner gets scrolled past and her clip gets watched and clicked. And the whole factory runs on a Claude subscription and a couple of MCP connectors and one laptop.

He closed the first brand in one evening: by morning the client had 30 ads with her face and copy for each and the first test batch pulled leads cheaper than his old agency. A week later the system shipped the next 30 while competitors were still loading a brief onto a designer.

Pause at 0:53: there go those 30 ads hitting the screen all at once

The comments fill with "so agencies are just not needed anymore" and that argument is the fuel the algorithm runs on.

Agencies are still writing briefs and waiting two weeks on a designer. And every Friday a fresh batch ships from him while the old world is still approving its first.

Rewind to 1:34 yourself: the command to launch all of this every Friday read it again

No studio. No team. No dollar on blind tests.

Just other people's ads as raw material. One machine instead of a department. And the discipline to scout the market first and generate second.

Half of you are already typing that this kills creative agencies. The other half are already opening the Meta Ad Library with a miner's eyes.

Which half are you in?

[Shade @twoshados]

https://t.co/xQYSGljVDr