My media project, «Mimyakish Workshop, ” is a channel documenting the entire journey—from 3D model to revenue earned from a collectible figure at an art market. I showcase the entire creation and sales process: how I model, print using a photopolymer printer, paint, package, transport to the market, and sell. I document this with real figures, prices, mistakes, and failures because it is this specificity that differentiates genuine trust from mere interest.

The concept is built on three elements that my direct competitors lack. First is the complete cycle instead of mere fragments, because competitors show isolated stages, whereas I illustrate the end-to-end path where every step connects to the previous one and leads directly to a sale. Second are real figures instead of vague claims; I openly disclose the cost of resins and equipment, and how much time is spent on printing and painting. Third is the Russian art market as a specific sales channel. I demonstrate how to prepare a booth, how to interact with customers, and what items sell well versus what doesn’t sell at all.

The tone of the media is calm, highly technical, and authentically lived-in—the voice of a craftsman in his workshop, not a guru pretending to be one. My primary platforms are TikTok, used to attract audiences through short process videos, and Telegram, used for retention through a daily log, financial updates, and drop announcements. Together, they create a complete funnel where TikTok brings in new people, and Telegram converts them into a loyal community and paying customers.

The first and most crucial source is my personal production log. I maintain a spreadsheet where I record every 3D-printed figure. This log includes the printing date, the model, the amount of resin used in grams, print time, post-processing time, materials consumed (primer, paint, varnish, packaging), and, of course, the final cost calculated by AI. This spreadsheet is the foundation for posts about economics, accurate figures, and pricing.

The second source of data comes from art markets. After every event, I record how many figures I brought, how many I sold, exactly which models moved, which ones remained, and at what price. Expenses for venue rental, travel, and packaging are included here as well. This data is necessary for honest analytical posts, such as «what didn’t sell and why.»

The third source consists of technical logs and screenshots from working software. From the slicer, I pull print settings, print time, and the number of layers. From ZBrush, I gather screenshots of the modeling stages. From online stores and receipts, I take current pricing for resin, isopropyl alcohol, paints, and brushes. All of this is transformed into content about material selection and beginner mistakes.

The fourth source is feedback from the audience via Telegram and TikTok. I collect comments, questions, poll results, and view statistics. I observe which topics generate the most discussion and what questions people want answered. This helps me adjust the thematic plan and understand what the audience is genuinely interested in.

The fifth source involves information about events—art market calendars, participation guidelines, space fees, and application deadlines. This is necessary for planning content vlogs and announcements in advance.

The human processing begins the moment I finish a print job or return from a market. I personally enter the preliminary data into my spreadsheets, noting exactly which model was printed, how many grams of resin were used, what paints and materials I utilized, and how long the painting took. I photograph my workstation, take screenshots from Blender and the slicer to document texture and detail, draft text for Telegram, formulate survey questions, and record short videos for TikTok. Ultimately, I make the decisions about what topic to pursue based on my experience and people’s reactions, and I verify all final figures and facts before publication.

Machine processing and neural networks engage as soon as I input the initial data or define the task. During the calculation phase, the neural networks and scripts automatically calculate the cost of each figurine by multiplying the resin and material consumption by current prices from receipts, determine the time spent by summing printing, post-processing, and painting hours, and calculate the net profit for each product unit and the overall profitability of the batch.In the planning stage, the neural networks analyze statistics from Telegram and TikTok—including views, retention rates, reactions, link clicks, and survey results. Based on these figures, the algorithms generate a posting schedule, advising on the most active times and days for the audience, and they also analyze comments, identify frequently asked questions, and suggest potential topics for future videos or posts.During the content generation phase, the neural networks write draft texts for Telegram based on my tables tracking expenses and process chronology. They can automatically assemble the structure of a post about a figurine’s cost, inserting the necessary figures, generate headline options, select hashtags, and even create text descriptions for videos. I then use these drafts as a foundation so I don’t have to start from scratch, but I always refine them as an editor—adding my own tone, emotion, and lively details.At the intersection of these two streams is my overall analytics, where the neural networks calculate percentages and build charts of income and expenses, which I then interpret to make decisions about scaling and determine what to tell the audience next month.

The first rule concerns fact-checking and verifying figures. Neural networks can make calculation errors or input incorrect data from tables, so I always double-check every number the system suggests before publishing a post. If the AI states that a figure cost one amount, but my table shows another, I correct it manually. Automation does not make financial decisions; it only provides a draft.

The second rule concerns speech recognition and subtitle generation. Automated subtitles on TikTok and Telegram often make mistakes with terminology, so I always proofread them before publishing. I make sure that words like «slicer,» «FEP film,» «isopropanol,» and paint names are spelled correctly.

The third rule concerns text generation. The neural network drafts the copy, but the final version always goes through me. I infuse it with my own tone, strip out bureaucratic jargon and templated phrases that often characterize machine-generated text. My posts should read like a master craftsman’s journal in the workshop, not a press release.

The fourth rule concerns hashtags and keywords. Platform auto-suggestions and neural networks can offer generic tags that are irrelevant. I only use those that accurately describe my content, such as 3D printing, miniature painting, art market, and handcrafted work. I don’t chase trending tags for reach if they are off-topic.

The fifth rule concerns auto-replies and bots. I do not have any bots in my Telegram that respond to messages on my behalf. Any automated reply must be labeled as an auto-reply, and I personally read every comment and message. People should receive a live response from me, not a template.

The sixth rule concerns automatic analytics. I use Telegram and TikTok’s built-in statistics and graphs generated by neural networks, but I don’t blindly trust the numbers because algorithms can inflate or deflate reach. Therefore, I look at the trend over a week, not at isolated spikes. If a graph shows a sharp jump, I check whether it’s related to a repost or a random feature in the recommendations.

The seventh rule pertains to automatic image compression. Telegram and TikTok automatically compress photos and videos, and sometimes this degrades the quality. If detail is crucial for a post—for example, a close-up of a paint finish or a model’s texture—I upload the file in high resolution and check how it appears after compression. If the quality drops significantly, I change the format or crop the image.

The eighth rule concerns confidentiality and data. I do not upload clients' personal information or the exact addresses of markets to neural networks or automatic services unless it has been agreed upon. I keep all financial spreadsheets locally, and I only upload anonymized summaries to the cloud. Neural networks must not have access to anything that could harm me or my customers.