How I Built Vireely, an AI Instagram Reels Analyzer for Uzbekistan

Summary

Vireely is a Telegram bot (@vireely_bot) that turns any public Instagram Reel into a clear report in Uzbek or Russian: the hook, the structure, the target audience, the lessons and ideas for your own content. I built it as an AI educator and product builder based in Tashkent, for creators and businesses in Uzbekistan.

In its first nine days, 129 people started the bot, 95 of them on a single day, and it delivered 45 full reel analyses. This post explains the thinking behind Vireely, the decisions that shaped it, what the first numbers taught me, and how the same approach can help anyone planning an AI product of this kind.

What problem does Vireely solve?

Many creators and small businesses in Uzbekistan copy successful Instagram Reels: the same music, the same format, the same trend. Most of the time the copy performs worse, because what made the original work was never the format. It was the hook in the first seconds, the order of the ideas, the promise to a specific audience.

That reasoning is invisible when you simply watch a reel. Most tools that analyse video content are built for English-speaking markets. Vireely was built to make the reasoning visible, in the languages people here actually use, so that a creator can learn from a good reel and then make an original one.

How does Vireely work?

1. The user sends a link to any public Instagram Reel to @vireely_bot in Telegram.

2. The bot downloads the public video. It never asks for an Instagram login, and it uses only official or public routes.

3. Google's Gemini model watches the reel as a whole: the speech, the text on screen and the picture.

4. The bot replies with a full report in Uzbek or Russian. On request it also records a short voice summary, so the user can listen instead of reading.

The whole interaction happens in Telegram, because that is where people in Uzbekistan already spend their day. There is no app to install and no account to create.

Why did I start from something that already worked?

The video analysis engine at the heart of Vireely already existed in an earlier bot I had built. Instead of starting from zero, I reused that engine and built a separate product around it. This saved weeks and meant the most difficult part was already tested on real reels.

The lesson came with a second one. At first the new voice summaries were drawing on the same AI quota as another of my projects, and they started to use it up. The fix was full separation: Vireely now has its own Google Cloud project, its own API key, its own quota, its own folder and its own data. Reusing code is smart; sharing limits and secrets between products is not.

Which product decisions shaped Vireely?

Every rule in Vireely exists for a specific reason. These are the decisions that mattered most:

- Three free analyses before paying. People need to see a real report on a reel they care about before they trust a product enough to pay for it.

- Charging only after the report is delivered, and only once per reel in each period. If an analysis fails, the user pays nothing, and asking about the same reel twice does not cost twice. Fair charging builds trust faster than any advertisement.

- Monthly plans and top-ups that never expire. Different users analyse very different numbers of reels, so the plans scale from a single creator to a small agency with several seats.

- Local payment methods. People in Uzbekistan pay with Click or by card transfer, so the bot accepts both.

- An admin panel with no password. The owner signs in with a one-time code sent to Telegram, so there is no password that can leak.

- One modest server for everything. The bot, its landing page vireely.uz and my personal website share one server until revenue justifies more. Infrastructure should grow with income, not ahead of it.

How did research replace guessing?

Before planning any content for my own accounts, I used the same engine to study what works for leading international creators in AI and business. The result is a research library of 92 analysed reels and 6 full profile breakdowns, about 114,600 words in total.

I then turned that library into a knowledge graph: a map of 978 concepts connected by 1,314 links and grouped into 76 topics. The graph shows which hook types appear again and again in successful content, such as myth-busting, before-and-after stories, proof with numbers, top-five lists, and fear or urgency.

Vireely also has a private Profile Lab. Every week it measures my own Instagram accounts against selected competitors and writes 30-day content plans in Uzbek. So far it has produced 8 content plans and 180 content ideas, each built on evidence from the research rather than on guesses.

What happened in the first nine days?

Vireely received its first user on 3 November 2026. By 11 November the numbers looked like this:

- 129 people started the bot.

- 95 of them arrived on a single day, 5 November.

- 45 full reel analyses were delivered, 4 of them instantly from the cache because someone had already asked about the same reel.

- 17 voice summaries were recorded.

- 5 users used their entire free trial.

- 3 users chose a paid plan.

- 12 users blocked the bot.

Read honestly, these numbers say two things. Interest is real: almost a hundred people arriving in one day shows that the problem resonates. Conversion is still early: about one person in forty-three has paid so far, which is normal for a product in its first week and is exactly what the next phase is designed to improve.

What mistake would I fix first?

The biggest lesson is about measurement. Of the 129 people who started the bot, only 10 arrived through tracked advertising links. The other 119 came from sources I cannot identify, mostly because the tracking was not in place before the first promotion. The counter for landing page visits was also added late.

This matters because the most important question for any young product is where the users come from, and I cannot fully answer it for the busiest day Vireely has had. Every link now goes through a tracked address that counts the click before opening the bot. If you are launching anything, set up source tracking before you post the first link, not after.

Why am I not selling hard yet?

Vireely is deliberately in a presence phase until mid-December 2026. My own audience is still small, and pushing sales to a small audience feels pushy and damages trust. For now, the content on my accounts is fully informative.

The plan is to sell first to the warm audience: people who already comment, send messages and follow a Telegram channel. Waiting for a large follower count before selling is a common mistake. A few hundred engaged people who trust you are worth more than many thousands who do not know you.

How should you think about building an AI product like this?

If you are considering a similar project, these are the principles I would keep:

1. Start from a behaviour you can see. People were already copying reels; Vireely helps them do it with understanding.

2. Reuse what already works, and isolate what must not be shared: keys, quotas, data and code.

3. Let people try the real thing before paying, and charge only for value actually delivered.

4. Fit local habits: the messenger people already use, the languages they speak and the payment methods they trust.

5. Measure where every user comes from from the first day.

6. Research before you create. Evidence beats intuition, especially in content.

7. Build trust before you sell, and sell first to the people who already know you.

8. Keep infrastructure small until the product earns the right to grow.

Where can this approach help a business?

The core idea behind Vireely is simple: an AI model watches or reads something and explains it in plain language. That idea applies far beyond Instagram.

A business could use the same approach to analyse competitors' video advertisements, review recorded sales calls, check product photos against a quality standard, or summarise customer feedback in Uzbek and Russian. Delivered through a Telegram bot, such a tool needs no app, no training and no new habits from the team.

The decisions described above, free trials, fair charging, local payments, separated keys and tracked links, transfer directly to any of these products.

Key takeaways

- Vireely is a Telegram bot that explains why an Instagram Reel works, in Uzbek or Russian.

- In nine days it reached 129 users and delivered 45 analyses, with 3 paid plans at the start.

- Reusing a working engine saved weeks; separating keys and quotas prevented one product from harming another.

- Free trials and charging only for delivered reports build trust.

- The biggest mistake was launching promotion before source tracking was ready.

- Research, not guessing, drives the content strategy, and selling waits until trust is built.

Do you need an AI product or a Telegram bot built?

I design and build AI products, Telegram bots and automations for creators and businesses, from the first plan to payments and analytics. If you want to see what this looks like in practice, try @vireely_bot in Telegram or visit vireely.uz.

If you are planning something similar for your own business, leave a comment under this post or contact me through this website. Describe the problem you want to solve, and I will tell you how I would approach it.