Voice & text control for Home Assistant using Telegram, Whisper & Gemini
📝 Workflow Description This workflow creates a conversational bridge between Telegram / n8n Chat and Home Assistant. It allows users to control smart home devices or request information using natural language (text or voice). ⸻ 🔑 Key Features Multi-channel input: Works with both Telegram and n8n’s chat interface. Voice support: Telegram voice messages are transcribed to text using OpenAI Whisper. AI-driven assistant: Google Gemini processes queries in natural language. Home Assistant integration: Uses MCP client tools to execute actions like turning devices on/off, adjusting lights, or broadcasting messages. Memory management: Short-term memory keeps context within conversations. Smart reply routing: Responses are automatically sent back to the correct channel (Telegram or chat). Message formatting: Telegram replies are beautified (bold, bullet points, inline code, links). ⸻ 📌 Node Overview Telegram Trigger: Captures incoming Telegram messages (text or voice). Bot Is Typing: Sends a “typing…” action to indicate the bot is working. Voice or Text: Separates voice and text inputs. Get Voice File → Speech to Text → Transcription to ChatInput: Handles Telegram voice notes by downloading the file, transcribing it, and preparing it for the chat pipeline. When Chat Message Received: Captures messages from n8n’s built-in chat interface. Process Messages: Normalizes incoming data (input text, source, session ID, voice flag). Home Agent: Main AI agent that processes queries. Google Gemini Chat Model: Language model for intent understanding and conversation. Simple Memory & Simple Memory1: Buffer memories to preserve conversation context. Home Assistant Connector: MCP client node that executes smart home actions (turn on/off devices, adjust lights, etc.). Reply Router: Routes the assistant’s response either to Telegram or to the n8n chat webhook. Telegram Message Beautifier → Telegram Send: Formats and sends responses back to Telegram. Respond to Webhook: Sends responses to n8n chat. ⸻ 🚀 Example Use Cases Send “Turn on the living room lights” via Telegram → Bot triggers Home Assistant action. Ask “What’s the temperature in the bedroom?” → Response comes back formatted in Telegram. Record a voice note “Goodnight mode” → Automatically transcribed and executed by Home Assistant. Use n8n chat to quickly trigger automations or check device statuses. ⸻ ⚡️ Benefits Unified chat & voice control for Home Assistant. AI-powered natural language understanding. Works seamlessly across platforms (Telegram & n8n chat). Extensible: new tools or intents can be added easily.
By ArkadiuszCreate celebrity selfie images and transition videos with GPT-4, SeedDream, and Kling
Who is this for? > *Trend-style celebrity selfie videos Content creators, social media managers, marketers, and AI enthusiasts who want to generate ultra-realistic AI selfie images with celebrities and create smooth cinematic transition videos. What this workflow does Part 1 - Image Generation: Collects your face image + up to 4 celebrity reference images via web form GPT-4 generates optimized prompts for realistic selfie composition SeedDream v4.5 AI creates photo-realistic selfie images All results automatically saved to Google Sheets Part 2 - Video Generation: Pairs generated selfie images (Celebrity 1 → Celebrity 2) Kling AI creates smooth walking transition videos Videos saved back to Google Sheets with metadata How to set up Configure API credentials (OpenAI, Fal.AI, Google Sheets OAuth2) Create a Google Sheet with two tabs: CelebritySelfies and CelebrityVideos Update the Config nodes with your Google Sheet ID Run! Requirements n8n instance (cloud or self-hosted) OpenAI API key with GPT-4 access Fal.AI account with API key Google account for Sheets storage Estimated costs ~$0.05 per image (OpenAI + SeedDream) ~$0.50-1.00 per video (Kling AI) 📧 Support For questions about this template : Website: bilsimaging.com Email: contact@bilsimaging.com
By Bilel Aroua