Timecode prep is slow, manual work

Before the first look is programmed, hours go into listening, marking and retyping. Light Helper AI does that first pass for you.

Listening and marking

You play the track again and again to find where the verse, chorus and breakdown start, then correct every marker by hand.

Counting bars

Cues look right when they hit the downbeat. Finding the “one” of every section by ear takes time, and it's easy to be a beat off.

Retyping on the desk

Then every cue name and every time has to be entered on the console. Light Helper AI gives you the list and the commands, ready to use.

01Desktop app·Main product

The desktop app

Everything the bot does, plus a timeline you can edit and a live link to the console. The Telegram bot is the free way to try the analysis first.

Waveform and timeline

Play the track and see beats, bars and sections on the waveform. Move a cue by hand when your ear disagrees with the analysis.

Live link to grandMA2 Beta

Connects to the console over Telnet on the local network and sends commands and macros directly. No files to carry over on a USB stick.

Finds consoles on the network

Scans the local subnet and mDNS for consoles and onPC stations, so you pick one from a list instead of typing IP addresses.

Encrypted console profiles

Address, port and login for each venue, stored encrypted on your computer. Switch between rigs in one click.

Macro export

Export the cue timeline as a grandMA XML macro for import, as plain command lines, as CSV, or copy it to the clipboard.

Windows, macOS, Linux

Builds for all three systems. The analysis engine is the same one the bot uses, so results match.

Free public beta for Windows, macOS and Linux. You sign in with your Telegram account. Found a bug or missing a feature? Write to the community chat.
02Telegram bot·Free now

Try the analysis free in Telegram

The bot runs the same analysis engine as the desktop app. Send a song and get the beat grid, bars, sections and a cue list with timecodes. No installation, no account.

L
Light Helper AIbot
♪ 02. Dreams.mp313 MB
✅ Track analyzed
Duration
5:36
BPM
111.0
Key
A Min
Sections
8
  1. 1Intro00:00:00:00
  2. 2Verse 100:00:17:15
  3. 3Chorus 100:00:43:13
  4. 4Verse 200:01:52:19
  5. 5Build00:02:27:06
  6. 6Chorus 200:03:06:04
  7. 7Verse 300:03:29:28
  8. 8Outro00:04:15:10
markers.csv commands.txt ma2_macro.xml analysis.json

Real bot output for “Dreams” by Romare, grandMA2 selected.

Telegram bot

Four steps in Telegram

No installation, no account. Works on your phone at the venue and on your laptop at the desk.

1

Pick your console

Press Start, choose the language, your console and firmware version, then the Timecode mode.

2

Send the track

MP3, WAV, FLAC, OGG, M4A or AAC, up to 20 MB. Send it as a file to keep the original quality.

3

Wait 20–60 seconds

The track is analyzed on our server. Nothing is filled in from a template: every number is measured from the audio.

4

Get your files

The section list arrives in the chat, followed by the cue list, console commands and the full analysis as files.

Telegram bot

What the analysis finds

Five things a programmer needs before building a timecode show.

Tempo and beat grid Works

Every beat of the track, accurate to one video frame (33 ms at 30 fps). The grid is corrected when the detector locks onto off-beat hi-hats, a common problem with house music.

Bars and downbeats Works

Where each bar starts, found from chord changes and the kick and bass. Section boundaries are moved onto the nearest downbeat, so cues land on the “one”.

Sections Works

Parts of the song are found by repetition. Material that comes back gets the same letter (A, B, C): the second verse is recognized as the verse again, the second chorus as the chorus.

Energy Works

Each section is rated low, medium or high relative to the rest of the track: a quick guide to where the show should build and where it should breathe.

Key Works

The musical key, measured from the harmony of the whole track, for example A minor. Handy when colour choices follow the mood of the music.

Same result every time

The analysis is deterministic: the same file always gives the same markers. No random numbers, no AI guessing about audio it cannot hear.

About section names. Boundaries and groups come from the audio. Names (Intro, Verse, Build, Chorus, Breakdown, Outro) are a suggestion based on position, repetition and loudness, so check them by ear. In electronic music the “Chorus” is usually the drop.
Telegram bot

Files for your console

The bot sends only the files that make sense for the console you selected.

markers.csv
WorksAll consoles

The cue list: cue number, section name, group, energy, start and end timecode, start bar. Opens in Excel or Google Sheets and imports into most DAWs.

Cue,Section,Group,Energy,Start TC,Bar
1,Intro,A,low,00:00:00:00,1
2,Verse 1,B,medium,00:00:17:15,9
3,Chorus 1,C,high,00:00:43:13,21
commands.txt
BetagrandMA2 · grandMA3

Command lines that create a sequence with one empty, labelled cue per section, plus the timecode of every cue for entering timecode events.

ClearAll
Store Sequence 101 Cue 3
Label Sequence 101 Cue 3 "Chorus 1"
#  Cue 3   00:00:43:13  Chorus 1  bar 21
ma2_macro.xml
BetagrandMA2

The same commands as a grandMA2 macro in the MA2 XML import format. Copy it to the importexport folder, import it into the macro pool and run it once.

Import "02.Dreams_ma2_macro" At Macro 1
Go Macro 1
analysis.json
WorksFor your own tools

Everything in one file: every beat and bar with its timecode, the sections, energy and key. Useful for scripts, plugins and visualizers.

{ "bpm": 111.0, "key": "A Min",
  "beats": [{ "beat": 1, "bar": 1,
    "downbeat": true, "timecode": "00:00:00:06" }, …] }
Why beta. The commands use only basic syntax (ClearAll, Store, Label) and are being verified on grandMA2 and grandMA3 onPC. Use a free sequence number: the default is 101.
Telegram bot

Ask about console commands in plain language

Switch the bot to Consultation and ask like you would ask a colleague. Answers take your console and firmware version into account and include command-line syntax.

  • “How do I copy cue 5 into another sequence on MA2?”
  • “What's the difference between presets on MA2 and MA3?”
  • “How do I set up a timecode show on grandMA3?”
Beta. Answers come from large language models and can be wrong. Try commands on onPC before using them in a show.
L
Light Helper AIgrandMA2 · v3.9
How do I name cue 3 in sequence 101?
Use the Label command:
Label Sequence 101 Cue 3 "Chorus"

Or press Label on the console and tap the cue in the sequence sheet, then type the name.
Telegram bot

Supported consoles

Pick your console and firmware in the bot. If your version isn't listed, enter it manually.

grandMA2

v3.7 · v3.8 · v3.9
  • Timecode filesCSV · TXT · XML
  • AI reference✓

grandMA3

v1.8 · v1.9 · v2.0
  • Timecode filesCSV · TXT
  • AI reference✓

Hog 4

v3.x · v4.x
  • Timecode filesCSV
  • AI reference✓

Avolites

Titan v12 · v14 · v19
  • Timecode filesplanned
  • AI reference✓

Bot or desktop app: what is the difference?

The bot is the free way in: send a track, get the markers and files. The desktop app is the full tool for building the show: edit the timeline, connect to the console and send cues directly. Both use the same analysis engine.

BotDesktop
Track analysis and cue list with timecodes✓✓
Files for grandMA2, grandMA3, Hog 4, Avolites✓✓
Edit cues on a timeline—✓
Send to the console over the network—✓ grandMA2
AvailabilityFree nowFree beta

What works today

Better you read it here than find out halfway through programming.

FeatureStatusNote
Track analysis: tempo, beats, bars, sections, energy, keyWorksTested on tracks with a known structure
Cue list with timecodes (CSV)WorksAny console
grandMA2 macro, grandMA2/3 command linesBetaImport being verified on onPC
AI reference for console commandsBetaCheck answers before a show
Automatic timecode events for grandMA2In developmentThe whole timecode show in one import
Show file conversion grandMA2 → grandMA3In developmentNot ready for real shows
AI effect generation for your fixturesIn development—
Desktop app: timeline editor, live link to grandMA2, macro exportBetaMain product · free beta, download

Current limitations

  • 4/4 only. Waltzes and odd meters get wrong bar numbers.
  • 30 fps. Timecode is generated at 30 frames per second; 24 and 25 fps are planned.
  • Bars, not phrases. Boundaries always sit on a downbeat, but not always at the start of an 8-bar phrase.
  • 20 MB per file. About 8 minutes of MP3 at 320 kbps, or 20 minutes at 128 kbps.
  • Bot languages. The bot speaks English and Russian. This site is also in Ukrainian and Spanish.

Next up

  1. First public release of the desktop app: timeline editor and live link to grandMA2.
  2. Automatic timecode events for grandMA2: sequence, cues and the timecode show in one import.
  3. Timecode export for grandMA3.
  4. Frame rate choice: 24, 25 or 30 fps.
  5. Phrase-aware sections and correcting section names right in the chat.

Questions

Is it free?

Yes. Everything is free during the open beta. If that ever changes, it will be announced in advance.

Do I need to install anything?

Not for the bot: everything happens in Telegram, on your phone or computer, and the console files are plain text, CSV and XML. The desktop app will be installed on Windows, macOS or Linux once it is released.

How accurate is it?

On music with a clear beat, beats are placed within one video frame. Sections are found by repetition and loudness: on most pop, rock and electronic tracks the boundaries are where you would put them yourself. Always listen through before the show.

What if a section name is wrong?

Names are suggestions; the times are what matter. Rename the cue on the console or in the CSV. If a boundary itself is clearly in the wrong place, send the track to the chat: that helps improve the analysis.

What about ambient music or tracks without drums?

If there is no steady beat, the bot says so and shows BPM as not found. Sections are still detected from the sound, but there is no beat grid.

The times are all shifted by the same amount. Why?

The file played at the show starts differently (extra silence, a fade-in, another edit). Analyze exactly the file that will be played, or offset the timecode start on the console.

What happens to my files?

Audio is processed on our server and stored with your project. It is not shared with anyone and is not sent to AI services. Questions in Consultation mode are sent to AI providers to generate the answer. Ask in the chat and we will delete your data.

I found a bug or the macro didn't import.

Write to the chat with your console, firmware version and the file. In beta this is exactly the feedback that matters most.

Try it on your next show

Download the desktop app, or check the analysis on one track in the bot without installing anything.