AI Video
& Images

Video and images for social media eat evenings: subtitles by hand, a cover in an editor, an explainer at a studio. Off-the-shelf AI tools give raw output: subtitles drift, iPhone colours burn out, a generated clip is eight seconds at most. I built five pipelines where AI generates and code assembles, checks and brings it to publishing

I use them myself and for a clinic owner: 22 clips went through subtitles, the designer bot made 296 covers, 4K images and short clips with sound were generated for the clinic, this website carries 40 thirty-second explainer videos rendered from code without shooting, and the voice assistant has a live face whose lips follow the voice

For a similar solution for your business, see AI Content Factory

What You Get

If you regularly need covers, short clips, a video explaining a service or subtitles, the pipelines do it from a word in Telegram. Paid generation runs only after your “go”, and publishing only on your button

How It Works

Drop a phone video in a folder and write “subtitles”. Speech is recognised with a time for every word, typos are fixed without moving the timing, up to four words on screen on a dark box, the spoken word in red. iPhone footage keeps its colours. About two minutes per clip

Write to the designer bot in Telegram what you need: a cover, a carousel, a portrait in your own likeness. The bot generates and returns the image in the chat, an edit is the next message. For the clinic — 4K images for posts

The bot assembles the prompt and generates a vertical clip of 4, 6 or 8 seconds with sound, optionally from a first and last frame. Every generation costs money, so it starts only after your “go”

Scenes are written as text with animation and rendered to video: wide and vertical, in two languages. No shooting, no voice-over. A text change is a rebuild in minutes, not a new edit. That is how the 40 videos for 10 services on this site were made

In the voice assistant a face comes up in about three seconds, its lips follow the voice, and when you interrupt, sound and lips stop together. It switches on with a button mid-conversation

A caption, hashtags and a title are drafted from the speech, and a card with platform switches waits for your button. Nothing is published without a tap

Tools

• Claude

  scripts, text fixes, captions

• Whisper and ffmpeg

  word-level speech, subtitles, colour, assembly

• Gemini and Veo

  images and short clips

• HyperFrames

  explainer videos from code

• Simli and LiveKit

  live face in the voice assistant

• Telegram and Google Drive

  commands, cards, files

Result

Subtitles, covers, short clips, service explainers and a live face are made from a word, not in an editor at night. Time saved I have not measured and do not promise. Not offered: an avatar with your face in a rendered video, or voice cloning

AI generates, code checks,
and money and publishing — only on your word

I want the same