Telegram video note bot

Storona A: a bot that turns a track into a video note with a spinning record

A demo bot: send a track and get a video note where a record with the track cover spins on a turntable. It is free and keeps no music.

Open the project Discuss a project

The task

A musician or a label wants people to play the track right in the chat and forward it on. A video note with a record and the cover plays at once, with no music player to open.

What we built

Audio, voice messages and mp3, m4a, flac, ogg and wav files up to 50 MB
The cover from the file tags, your own photo or a label made from the title
The loudest part, the start, the middle or your own range of up to a minute
Three vinyls and an optional tonearm, a one-minute note in about 3.5 seconds

A video note instead of a link

Storona A turns a song into a Telegram video note. A record spins on a turntable, the track cover sits in the middle of the label, and the chosen minute plays. The bot is live and free: find @ashehaipulsebot in Telegram and send it any track. The brand was made up for the demo, and the bot sells nothing.

Which minute goes into the note

A Telegram video note lasts up to a minute. A short track goes in whole, and from a long one you pick a part: the start, the middle, your own range like 1:23-2:05 or the loudest part. The bot finds the loudest part by the average loudness of half-second blocks, and it usually lands on the chorus. The sound fades in and out over half a second.

Cover and vinyl

The bot takes the cover from the file tags. If there is none, it takes the thumbnail from Telegram. If that is missing too, it asks for a photo, and without a photo it draws a label with the title and the artist.
The “Own cover” button puts any photo of yours in the middle of the record.
There are three vinyls: black, clear red and marble, whose swirls take the colours of the cover. The tonearm can be removed.

Why a note takes seconds

Pillow and NumPy draw the record, the glare and the tonearm, and ffmpeg builds the video. One turn of the record takes 3.6 seconds, which is 108 frames. ffmpeg renders only those and then repeats the finished turn, so a one-minute note takes about 3.5 seconds in the bot’s Docker image. When many tracks arrive at once, they queue up, and everyone sees how many jobs are ahead.

The bot has no backend because there is nothing to store. The track, the cover and the finished video sit in a temporary folder while the note is built and are deleted right after sending. If you need a bot that makes video notes from video, music or photos in your own style, see the video note bot for Telegram page. The chat screenshots are mockups built from the bot’s real texts, while the records and the note in the video were rendered by the bot itself.