Project Title: Global AI Media Bot (Telegram + Make.com)Type: Automated service for generating personalized videos.Platform: Make.com (Integromat) + API.
1. Project Description
It is necessary to develop a Telegram bot that generates personalized content (predictions/motivation) based on the user's request.Key Feature: The user sends a request (by voice or text) — the system returns a vertical video (15-30 sec) with unique visuals, voiceover, and subtitles.Geography: Worldwide (Multilingual).Monetization: Telegram Stars (XTR) + Credit system.
2. Technical Stack (Architecture)
All logic is built on Make.com. Scenarios (Blueprints) must be transferred to the client's account after completion.
Required Integrations (API):
Frontend: Telegram Bot API.
Logic & Text: OpenAI API (gpt-4o or gpt-4o-mini).
Transcribing: OpenAI Whisper (for translating voice messages into text).
Image Generation: Replicate API (Model: flux-schnell or sdxl-lightning).
Voice Generation: ElevenLabs API (Model: eleven_multilingual_v2 — mandatory for language support).
Video Rendering: Creatomate API (MP4 assembly: Image + Zoom + Overlay + Audio + Auto-subtitles).
Database: Google Sheets or Airtable (for user and credit accounting).
3. Database and Credit System
User profiles must be stored in the database.Fields: user_id, language, credits_balance, registration_date.
Credit Logic:
4. Localization (Auto-Detection)
The bot automatically determines the language without asking the user at the start.
Algorithm:
Check the Deep Link parameter (t.me /bot?start=en -> EN).
If there is no Deep Link, check message.from.language_code from Telegram.
Save the language in the variable {{lang}}. All subsequent responses (text and voice) are generated in this language.
5. Payment System (Telegram Stars)
Method: sendInvoice (Currency: XTR).Products (packages):
1 Video Session (for example, 350 Stars) -> Grants +1 credit.
Package "Trinity" (for example, 1000 Stars) -> Grants +3 credits.
Package "Maximum" (for example, 1750 Stars) -> Grants +5 credits.
Requirement: Implement answerPreCheckoutQuery and processing of successful_payment, which instantly updates the balance in the database (Google Sheets).
6. User Flow
Stage A: Onboarding & Free Tier (Free)
Start: The user presses /start.
Survey: The bot asks 3 questions with buttons (Name / Date of Birth / Area of Inquiry).
Free Result: The bot generates a "Card of the Day".
Upsell: The bot offers: "To receive a personalized Video Vision and answer to your question — purchase a full session."
Stage B: Paid Tier (Video Generation)
The user pays for a Stars package (or uses existing credit).
The user sends a request (Voice or Text).
Processing in Make.com:
Whisper: Transcribes voice.
GPT-4o:
Generates text response in the user's language ({{lang}}).
Generates a prompt for the image in English (Style: Cinematic, Abstract, Mystical, No Text, No Faces).
Replicate: Generates an image (9:16).
ElevenLabs: Voices the text response.
Creatomate: Renders the video (Slow Zoom on the image + Overlay effects + Subtitles).
Delivery: The bot sends the finished MP4 video.
7. Quality and Stability Requirements
Stylistics: Modern Dark Fantasy / Abstract style. No "crooked" people and fingers in images (use negative prompts).
Error Handling: If the API (for example, Replicate) returns an error or times out, the bot must inform the user "Connection interrupted, I will try again" and make an automatic retry of the request, rather than "crash".
Admin Panel: Ability to change prices (number of Stars) and system prompts through variables in Make.com, without rewriting the entire code.
8. Deliverables
The contractor delivers:
A fully configured bot.
Scenarios (Blueprints) in Make.com.
A configured database table.
A video template in Creatomate.
A short video instruction on management (how to change the price, how to change the prompt).
Deadline: up to 7 days.Budget: By agreement.
