AI Speech Coach and Presentation Generator via Telegram - n8n Workflow

• Professional speakers and presenters needing instantaneous feedback on draft speeches.


• Content creators looking for an efficient way to iterate on spoken content.


• Users of n8n seeking a practical example of integrating Telegram, OpenAI, and Google Gemini into a conversational agent.


• Developers interested in complex n8n templates featuring dynamic prompting and memory management.

Developing a compelling speech often requires iterative drafting and expert feedback. This advanced n8n workflow solves this by creating a personal, multimodal AI coach accessible directly through Telegram. The user can simply send a voice note or type their draft. The n8n node structure intelligently routes the message: voice messages are transcribed using OpenAI’s Whisper, and text messages proceed directly. The core of this n8n template is the dynamic prompting mechanism, which uses the Google Gemini LLM to provide feedback, guide speech preparation, or generate a complete speech based on accumulated context. This ensures the output is always relevant, making this n8n workflow an indispensable tool for public speaking preparation.

This comprehensive n8n workflow begins with the Telegram Trigger n8n node, awaiting an incoming message (the initial n8n trigger).

1. Input Processing: The workflow first checks if the input is a text message or a voice message. If it is a voice message, the flow uses a Telegram n8n node to download the file, followed by an OpenAI n8n node (LangChain integration) to transcribe the audio into text (Whisper). If it is a text message, it proceeds immediately.


2. Intent Routing: The Route Flow Based on Message Content n8n node (a Switch) analyzes the text for keywords like 'new speech' or 'generate speech'.


3. Dynamic Prompting & Memory Management: Based on the intent, one of three system prompts is dynamically injected. If the user requests a 'new speech', the Wipe Conversation Memory n8n node is executed first to ensure the AI starts fresh, minimizing hallucinations. The ongoing conversation is stored using the Store Conversation Memory n8n node.


4. AI Generation: The AI Agent n8n node utilizes the defined prompt, conversation memory, and the powerful Google Gemini Chat Model to generate the specific feedback or speech draft.


5. Output Preparation: Since Telegram has character limitations and formatting quirks, two Python Code n8n nodes are used sequentially: one cleans the response by removing unwanted Markdown characters, and the second splits the long AI output into chunks of less than 4000 characters.


6. Final Delivery: The workflow completes by sending the chunked, cleaned responses back to the user via the Respond to Telegram Message n8n node.

To deploy this powerful n8n template, follow these steps:

1. Import: Copy the provided n8n workflow JSON and import it into your n8n instance.


2. Credentials Setup: This n8n workflow requires three distinct credentials:

1. Webhook Activation: Ensure the main Recieve Telegram Message n8n trigger is active to start listening for incoming messages from your Telegram bot.


2. Test: Send a test message (text or voice) to your Telegram bot to confirm the entire n8n workflow executes correctly, from transcription through AI response and output chunking.

Recieve Telegram Message (n8n trigger): The entry point of this n8n workflow. It listens for new message updates from the configured Telegram bot, initiating the entire process.
If Voice Message (n8n node): A conditional n8n node that checks if the incoming message contains text. If text is absent, the flow branches to download and transcribe the voice note.
Transcribe Audio File (OpenAI n8n node): Utilizes OpenAI's Whisper model (via the LangChain n8n node) to convert the downloaded audio file into usable text.
Route Flow Based on Message Content (Switch n8n node): The key routing mechanism. It examines the user's text input to determine if the intent is to start a 'new speech' (clearing memory) or 'generate speech' (synthesis) or request general 'feedback' (default path).
Wipe Conversation Memory (MemoryManager n8n node): Crucial for context reset. This n8n node deletes all past interactions stored for the specific user's chat ID when a new topic is initiated.
Google Gemini Chat Model: Provides the language model used by the AI Agent. Configured here as models/gemini-2.0-flash-001 for speed and efficiency.
AI Agent (LangChain n8n node): The core decision and response engine, pulling the dynamic system prompt, the user's input, and the stored memory to generate an optimized response.
Code to split output into chunks under 4000 characters (Code n8n node): A Python script designed to handle Telegram's message size limit, ensuring long AI-generated speeches are delivered sequentially in manageable chunks.