Veo-3: AI Video With Sound in Telegram

You have an idea for a short clip with sound, but no camera, no actors and no editor. Veo-3 for Telegram takes a text description and returns a video without leaving the messenger. 🎬

This is an unofficial mini app built around the Veo 3 model, which Google DeepMind released in May 2025. The bot itself is run by independent developers.

Picture and sound in one go

The big idea behind Veo 3 is that video and audio are generated together. A clip can contain spoken lines, sound effects and background ambience, so you skip the dubbing step.

Clips from this kind of model are short. Veo documentation usually lists 4, 6 or 8 seconds. For anything longer you stitch several fragments yourself.

Your first clip in four steps

1. Open the bot and press /start. The mini app loads inside Telegram.
2. Pick a language with /lang if the interface opened in the wrong one.
3. Describe the scene: who is in frame, what happens, what the light and sound are like.
4. Wait for the render and save the video to your device.

Prompts that get better results

The model rewards specifics. Instead of "a beautiful city", write: "a night street after rain, neon reflected in puddles, a saxophone playing far away, the camera slowly moving forward".

• One shot per clip. Eight seconds cannot hold a story with five locations.
• Camera movement. Say whether it flies, stays still or pushes in.
• Sound. For speech, put the line in quotes and describe the voice.
• Style. Film, cartoon, documentary: choose one.

Subscription and cost

The command list includes /cancelvip, so access works as a subscription that you can cancel inside the bot. The terms are in the offer document linked from the bot profile. Read it fully before paying: it covers billing periods, refunds and usage rules.

Generating video from scratch is expensive even for the model's own maker. Reviews of large providers put the price of a second of video with audio at tens of cents, so do not expect a free unlimited service.

Support is available through /support. If a clip fails or you were charged twice, write there with your prompt and the payment time.

Copyright and ethics, briefly

• Real people. Do not make clips where a recognizable person says things they never said. That can mislead viewers and break the law.
• Brands and characters. Using someone else's brand or character in an ad without permission is risky.
• Labeling. When you post a clip online, say it was generated by a neural network.

The app has a small user base and is independent, so for important projects keep your finished files locally instead of relying on storage inside the bot.