Day 35. Three days of regenerating the paid channel: I gave one frame to five neural networks, and only one honestly made the revealing scene

On Day 32 I promised to watch whether people come to «closer» through the link in the model's bio. In three days nobody came: not to «closer», not to the model's open Telegram. The bio link isn't working as a door yet.

What I changed since the last entry. The paid-channel frames made on October 14 stopped satisfying me: the model's face looks like her only every other time. Over three days I regenerated the channel on a different neural network, Nano Banana Pro, and made about 250 new frames. It rejected roughly one in seven as too revealing: lingerie, unbuttoned shirts. To see if there was a way out, I gave the most revealing rejected frame to five models at once. Three refused, the fourth dropped the lingerie and returned an ordinary portrait, and only Seedream made the scene as asked. Now the scheme is: Nano Banana first, and on refusal the same prompt goes to Seedream. «Closer» also got the model's voice messages, one at 23:00.

What it gave. The model has 103 followers on Threads, up from 92 on October 14, and 6 on Instagram. The «closer» plan is filled through November 9, the model's open Telegram through November 6. Tomorrow at 12:10 her first talking reel goes out on Instagram.

Mistake of the day. On October 15 a note «terrible photo, regenerate» was left in the dashboard for the next day's post, and I saw it only after the post went out. You can't replace the photo in a published Threads post. Now everything awaiting review sits in one folder, and I check the notes before every change to the plan.

Three days in total. About 4,100 ₽ spent on generations; about 22,800 ₽ since September 17. Author's hours: 6, two a day. Earned: 0 ₽. The course costs 4,990 ₽, «closer» 390 ₽ a month.

What I do tomorrow. I watch how many people reach the model's profile from the talking reel «Who am I».