Emocute Studio Guide

A step-by-step walkthrough for first-time users: installing, making your first track, handing it to Suno, and fixing common problems. You don't need to be comfortable with the terminal (the window where you control your computer by typing).

Your computer
Your AI

Only the steps for the computer and AI you selected are shown.

1What it is

Emocute Studio lets you control the chord progression of a Suno track just by talking to an AI.

  1. Say something like "a wistful lo-fi track"
  2. The AI designs the progression and gives you a preview MP3 plus the text to paste into Suno
  3. You hand that MP3 and text to Suno
  4. Suno builds the song on the progression you aimed for

No music theory needed. If you do know theory, you can hand it the exact chords you want.

It is not a standalone app. You open the downloaded folder with an AI (Claude Code or Codex) and work from there.

If you have the MIDI edition, skip the Suno parts (MP3, handing to Suno, Style). Play the exported MIDI in your DAW instead.

2What you need

Computer
Mac or Windows
AI (pick one)
Claude Code (a paid Claude plan) or Codex (a paid ChatGPT plan; tested on Plus)
Suno
Pro or Premier recommended (needed for Upload Audio, which sends the MP3 to Suno)
Internet
For the first setup (including a ~206 MB sound font) and for updates

Python and the other small parts are installed by the AI on first run. You don't need to gather them yourself.

3Getting started (3 steps)

If this is your first time, follow these in order.

1. Install the AI

Open Terminal (search "Terminal" in Launchpad). Paste this line and press Enter.

Type in the terminal
curl -fsSL https://claude.ai/install.sh | bash

Once installed, close and reopen the terminal, type claude and press Enter. The first time, a browser opens so you can sign in with your Claude account.

Official Claude Code install guide →

2. Unzip, then start the AI inside that folder

After purchase, download the ZIP from your library on emocutelab.com and unzip it. The English version is the folder whose name ends in _en. You can put it anywhere.

  1. In Terminal, type cd followed by one space (don't press Enter yet)
  2. Drag the unzipped folder onto the Terminal window and let go (its location is filled in for you)
  3. Press Enter

In the window that opens, type this and press Enter.

Type in the terminal
claude

The first time you open the folder, it asks whether it may work in this folder. Allow it.

Always start it inside the unzipped folder. Started anywhere else, the AI can't load Emocute Studio's rules.

3. Say "set it up"

Say to the AI

set it up

The AI installs the remaining parts one by one (Python packages, FluidSynth, ffmpeg, the sound font). It only asks you to step in where it can't click for you, such as typing a password or going through an installer.

When it replies that setup is done, you're ready. Just tell it the mood of the track you want.

If setup created a .venv (a parts folder just for this project), type source .venv/bin/activate before launching each time from then on. The AI will tell you when this applies.

4Your first track

In the window you started, just describe the mood you want.

Say to the AI

Suggest a few tracks that feel good and have a surprise in them

  1. The AI offers a few ideas (each with a genre, mood and chord direction).
  2. Reply with the number you like (e.g. 2).
  3. The AI shows a short draft of eight lines or fewer and waits. Say "go" if it's right, or tell it what's off.
  4. Once the draft is approved, it goes all the way through: lyrics, MP3 and the text for Suno.

You can also be specific.

Say to the AI

Make a footwork track. Key of Fm, 160 BPM.

Say "do it all without checking with me" to skip the draft and go straight to the end.

What you get

Three things appear in the renders folder inside the unzipped folder.

Preview MP3
The progression played on keys and pads only. This is what you hand to Suno (called the "bed")
Prompt HTML
Open it in a browser to see the Style, Exclude, lyrics and settings to paste into Suno
MIDI
One .mid per part, for playing in a DAW (optional)

5Handing it to Suno

This is the most important part: Suno does not read chord names as text. Writing "Fmaj9 - E7 - Am9" in the Style or lyrics field is ignored, and Suno falls back to generic chords. The progression only reaches Suno when you give it the preview MP3 as audio.

  1. Listen to the MP3 in renders and make sure you like the progression (to change something, tell the AI — see "Example requests" below)
  2. Open the prompt HTML in your browser
  3. In Suno's create screen, upload the MP3 (Upload Audio)
  4. Paste the Style, Exclude and lyrics from the HTML into Suno's fields of the same names
  5. Set the values in the table below and generate
  6. Generate a few takes and keep the best one

Suno settings

Model
v6
Audio Influence
Around 80 (how closely Suno follows the uploaded MP3)
Weirdness
80
Style Influence
80–100
Variety
Off (under More Options; on Normal, Suno rewrites your Style)

If takes keep missing, listen to the MP3's progression again before touching the Style wording. If it still doesn't fit, lower Audio Influence: Suno then uses the MP3 as a loose reference and fills in the rest.

6Example requests

Copy any of these and change the words. Loose descriptions work fine.

Ask by mood

Say to the AI

A wistful lo-fi track

Say to the AI

Dark verses, then open up in the chorus

Get particular about the progression

Say to the AI

F minor, 84 BPM, one chord per bar. Design an 8-chord progression with smooth voice leading and one tritone substitution. Resolve at the end.

Asking for eight or more chords, one or two reharmonisation tricks, and a resolution at the end gets you out of the usual four-chord loop.

Hand over the chords you want

Say to the AI

Use these chords, one per bar: Dbmaj9 - C7(b9) - Fm9 - Ebm9 - Ab13 - Dbmaj7#11 - Gm7b5 - C7alt

Emocute Studio handles the voicings and how each chord connects to the next.

Listen and fix

Say to the AI

The fifth chord sticks out. Replace it with one that flows more smoothly into the sixth

Say to the AI

It's too bright overall. Keep the movement, but make bars 3–4 sound darker

Name the spot clearly — "bar 3", "the fifth chord" — and the fix lands where you want it.

Use a track you have as a reference

Say to the AI

Analyse reference.mp3 and give me a Suno command

You get a rough key and BPM, how bright the sound is, and wording for Suno. It does not read the chord progression. Put the reference MP3 inside the unzipped folder.

Check a track Suno sent back

Say to the AI

Separate and analyse suno_returned.wav

It splits the track into vocals, drums, bass and other, and checks the sound. The first time needs an extra part — just ask the AI to install it.

Get feedback on lyrics

Say to the AI

Have another AI critique these lyrics

A separate AI from the one that wrote them scores the lyrics on six points, such as hook strength, and returns the weak lines with suggested fixes. It only runs when you ask.

7Starting new chats and usage

The longer a conversation gets, the more each exchange uses.

  • One track per chat is the norm. Once the draft is approved, let it run to the end in one go.
  • If redo after redo makes the conversation long, move to a new chat.
  • Before moving, ask it to "write a handoff". In the new chat, have it read that file and it picks up where you left off.

To start a new chat, close the terminal window and launch again as in step 2 of "Getting started".

With Claude Code, usage comes out of your Claude plan's allowance.

8Updating

When a new version is out, the AI tells you in one line when you start it.

Say to the AI

update

The latest version is unpacked as a new folder next to your current one. Your current folder is left untouched. When it's done, start the AI again in the new folder. The sound font is copied over automatically, so there's nothing to re-download.

First update only: register your key

  1. Sign in at emocutelab.com, open Emocute Studio in your library and press "Show auto-update key"
  2. In your own terminal, go into the unzipped folder (same as step 2 of "Getting started") and type the line below
  3. Paste the key and press Enter (the key isn't shown on screen)
  4. Tell the AI "update" again
Type in the terminal
python3 tools/self_update.py --set-key

Don't paste the key into the AI conversation, and don't share it with anyone.

If you own any v8 version, updates are free, as many times as you like.

Release notes →

9Troubleshooting

First, ask the AI "what should I do?" with the error still on screen. It usually fixes things itself. If you're still stuck, open the symptom that matches.

Typing claude (or codex) says "not found"

Either the AI isn't installed yet, or you haven't reopened the terminal since installing. Close the terminal, open it again and retry. If it still fails, redo step 1 of "Getting started".

The AI acts like it doesn't know Emocute Studio

It was started outside the unzipped folder. Start it again inside the folder, as in step 2 of "Getting started".

No MP3 is produced

FluidSynth or ffmpeg is missing. Tell the AI "set it up" again. If you get a WAV but no MP3, it's ffmpeg.

It says the sound font can't be fetched (missing_soundfont)

The first sound font download (~206 MB) needs a connection. Check you're online — and on Codex, allow network access — then ask again. A file cut off midway is fetched again automatically next time.

On Mac it stops with externally-managed-environment

This happens with Homebrew's Python. Ask the AI to "create a .venv and install there". From then on, type source .venv/bin/activate before launching each time.

Parts are installed but errors continue

You may have more than one Python, and the parts went into the other one. Ask the AI to "reinstall with python3 -m pip" (on Windows, python -m pip).

Codex says network access was blocked

Codex blocks network access for safety. Allow it when asked. If it doesn't ask, close Codex and reopen it with codex.

Codex on Windows doesn't write anything

The "Set up Agent sandbox to continue" setup on first launch wasn't completed. Reopen Codex and go through the setup.

It stopped because I wrote an artist name

Suno doesn't accept artist names, so Emocute Studio stops you first. Instead of "like [artist]", describe the sound (e.g. grimy Memphis phonk, muffled keys).

Suno doesn't follow the progression

Check that you uploaded the MP3 with Upload Audio. Chord names written as text never reach Suno. If you did and it still misses, check Audio Influence is around 80 and generate several takes. The fallback for when it still won't fit is at the end of "Handing it to Suno".

On Windows ARM, only reference analysis doesn't work

Some required parts don't support ARM Windows, so reference analysis and separation are unavailable there. Everything else — progressions, MP3, Suno text, lyric feedback — works.

It's slow, or the AI is reading lots of files

Tell it "only read what you need".

10Glossary

Terminal
The window where you control your computer by typing. "Terminal" on Mac, "PowerShell" on Windows
Claude Code
Anthropic's official tool for running its AI (Claude) inside the terminal
Codex
OpenAI's official tool for running its AI (ChatGPT) inside the terminal
Bed
An MP3 of the progression on keys and pads only, used to get the progression into Suno
Style
The Suno field that describes how the track sounds. Emocute Studio writes it in detail
Exclude
The Suno field for sounds you don't want
Audio Influence
How strongly Suno follows the uploaded MP3
MIDI
A file that stores only pitches and lengths. A DAW can play it with any sound you like
DAW
Music production software such as Ableton Live, Logic Pro or FL Studio
Reharmonisation
Swapping chords for different-sounding ones while keeping the same flow

11What you can do with it

  • Music you make is yours to use freely, including selling and streaming. No credit required.
  • You can modify the contents to suit your own setup.
  • You may not share, resell or publish Emocute Studio itself (the program and documents).

See the bundled LICENSE.md for details. Corrupted downloads, malfunctions and the like are fully refunded if you contact us within 14 days of purchase.

Refund policy →

12Contact

Questions or bugs: [email protected]. Including which AI you use (Claude Code / Codex), your computer (Mac / Windows), and the exact text on screen helps us answer faster.

← Emocute Studio