Generate AI videos from Telegram messages with Veo-3

Find Your AI Training Job
Chat with DigiNo Toucan and get matched to remote roles in one minute.

Social media clients always need more short-form video content but rarely have a fast way to go from idea to deliverable. This AI automation lets you offer a done-for-you pipeline where a client texts an idea and gets back a matching thumbnail and video clip within minutes.

What This Automation Does

Accepts a text or voice idea sent through Telegram and converts it into two structured prompts, one for a still image and one for a video clip
Generates a thumbnail image via KIE.ai using Gemini 2.5 Flash Image and delivers it to the client inside the same Telegram chat
Uses that thumbnail as the starting frame for a Veo-3 video generation job, keeping colours, objects and mood visually consistent between the two outputs
Delivers the finished MP4 clip back into Telegram once rendering completes, giving the client a ready-to-post thumbnail and video with no context switching

Tools Used

n8n
Telegram
OpenAI

Work faster

Stop Typing and Start Speaking With Typeless

Typeless turns what you say into clean, formatted text in any app on Mac, Windows, Linux, iOS and Android, which it rates at 4x faster than typing. It strips filler words, keeps only your final phrasing when you change your mind mid-sentence, and works in 100+ languages. Emails, AI prompts, client messages and docs, done by voice.

Try Typeless →

Where to Get Hired for This Skill

On Contra, top freelancers across this stack have earned 102 combined verified reviews from real client projects.

Source: Contra freelancer search · refreshed 30 May 2026

Start Earning as a Freelancer on Contra

Contra is a commission-free professional network for independents. Browse live AI automation work and keep what you earn.

Join Contra Free →

How To Build It

Connect Telegram as the idea intake channel

Wire up a Telegram bot so that every message a client sends, whether typed or a voice note, arrives as a trigger that kicks off the rest of the pipeline automatically.

Route voice and text inputs correctly

Add conditional logic so that voice messages are transcribed via OpenAI before being processed, while plain text messages skip that step, ensuring both input types produce a clean idea string.

Build structured prompts with an AI agent

Pass the cleaned idea text into an OpenAI-backed agent with a structured output parser that produces two separate, optimised prompts: one describing the thumbnail image and one describing the video motion and scene.

Generate and deliver the thumbnail first

Send the image prompt to the KIE.ai Nano Banana endpoint, poll for completion, and push the returned PNG back to the client in Telegram before the video job even finishes, so they see progress immediately.

Launch Veo-3 video job using the thumbnail as init image

Submit the video prompt to the KIE.ai Veo-3 endpoint with the thumbnail attached as the initial frame, poll until the render is complete, then deliver the MP4 clip directly into the same Telegram conversation.

Pitfalls

KIE.ai video render times vary significantly under load, and the polling interval you set can either flood the API with requests and trigger rate limits or leave the client waiting much longer than expected if the interval is too generous.
Veo-3 requires the init image to meet specific resolution and aspect ratio constraints, so if the thumbnail generation step returns an image with unexpected dimensions the video job will fail silently and you will need error-handling logic to catch and retry it.
Telegram bot tokens expire or get revoked if the bot is inactive for extended periods or if a client accidentally restarts BotFather setup, breaking the trigger without any obvious error until the next time a message is sent.

FAQ

Can I build this without coding?

Yes. The entire workflow is built inside n8n using a visual interface and pre-built nodes. The only configuration that requires care is correctly formatting the API calls to KIE.ai, which involves setting headers and polling logic, but no custom code is required to do that.

How long does it take to build?

A working prototype typically comes together in two to four hours. Getting the structured prompt logic reliable and the polling intervals tuned for production usually adds another session on top of that.

What can I charge for this service?

Pricing is entirely your decision based on your market and client relationships. The workflow is well-suited to a monthly content retainer model where a fixed number of videos are delivered each month, or a per-clip delivery model for clients with irregular volume.

Which tools are required versus optional?

Telegram, KIE.ai and n8n are required as the core input, generation and orchestration layers. OpenAI is used for voice transcription and prompt structuring but could be swapped for another LLM provider if you already have a preferred integration, making it technically optional with some rework.

This is original DigiNo analysis. The underlying automation pattern is a community workflow template – view the original on n8n.