MiniMax Music 3 identity tournament

A controlled singer-identity comparison across Serj-only and Daron-only MiniMax Music 3 language-model LoRAs. It also compares random 128-frame training windows against continuation windows while holding the dataset split, instrumental regularisation, optimizer, prompts, inference seed, and LoRA-strength sweep fixed.

This card cross-links 640 FLAC renders hosted by the four source repositories. Audio is not duplicated here.

Design

leg singer tags window objective instrumental regularisation checkpoints
Serj, random window Serj Tankian random 128-frame crop 50% singer-matched 10
Daron, random window Daron Malakian random 128-frame crop 50% singer-matched 10
Serj, continuation window Serj Tankian continuation, 128-frame target 50% singer-matched 10
Daron, continuation window Daron Malakian continuation, 128-frame target 50% singer-matched 10

Machine-readable comparison manifest

Singer identity and training-window comparison (640 renders)

Serj-only and Daron-only LoRAs trained with random or continuation windows, each with singer-matched instrumental regularisation.

All renders use a 40-second maximum duration, 30 flow steps, flow CFG 1.7, and seed 20260824. MiniMax Music 3 may end a song before the duration cap.

Training legs

No trigger word / off-genre descriptors

Caption: Alternative pop-jazz, 102 BPM, brushed drums, upright bass, clean electric piano, airy male vocals, smooth verse-chorus songwriting, restrained dynamics, polished intimate mix.

LoRA strength 0.25
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.50
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.75
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 1.00
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window

Trigger word only

Caption: system of a down, soad style music

LoRA strength 0.25
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.50
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.75
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 1.00
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window

Matching descriptors, no trigger word

Caption: Lilting hypnotic arpeggios, a gentle melodic mid-tempo groove with layered harmonies, sudden explosive distorted riffs, unstable meter changes, theatrical male vocals, tightly syncopated bass and drums, abrupt shifts between fragile verses and frantic heavy choruses.

LoRA strength 0.25
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.50
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.75
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 1.00
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window

Trigger word + matching descriptors

Caption: system of a down, soad style music, lilting hypnotic arpeggios, a gentle melodic mid-tempo groove with layered harmonies, sudden explosive distorted riffs, unstable meter changes, theatrical male vocals, tightly syncopated bass and drums, abrupt shifts between fragile verses and frantic heavy choruses.

LoRA strength 0.25
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.50
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 0.75
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window
LoRA strength 1.00
training leg step 500 step 1000 step 1500 step 2000 step 2500 step 3000 step 3500 step 4000 step 4500 step 5000
Serj, random window
Daron, random window
Serj, continuation window
Daron, continuation window

Evaluation lyrics

The lexical content is fixed; only the singer assignment in section tags changes.

Show Serj Tankian section tags
[Intro]
Count the teeth in the traffic light
Paint the silence ultraviolet

[Verse 1: Serj Tankian]
Paper crowns on the factory floor
Everybody wants a little more
Wind-up prophets in a checkout line
Selling borrowed seconds back as time

[Pre-Chorus: Serj Tankian]
Turn the dial, divide the sky
Ask the wires to tell us why

[Chorus: Serj Tankian]
Wake up the walls, let the ceiling shake
We built a feast from a plastic lake
Hands in the air while the engines stall
Who signed the order to swallow it all?

[Verse 2: Serj Tankian]
Velvet sirens in the avenue
Paint every warning a pleasant blue
I fed my shadow to the voting machine
It came back polished, quiet, and clean

[Bridge: Serj Tankian]
Softly, softly, count to four
Kick down the arithmetic door
One for the hunger, two for the show
Three for the ones who already know

[Final Chorus: Serj Tankian]
Wake up the walls, let the ceiling shake
We built a feast from a plastic lake
Hands in the air while the engines stall
Who signed the order to swallow it all?

[Outro]
Count the teeth
Kill the light
Carry the silence
Into the night
Show Daron Malakian section tags
[Intro]
Count the teeth in the traffic light
Paint the silence ultraviolet

[Verse 1: Daron Malakian]
Paper crowns on the factory floor
Everybody wants a little more
Wind-up prophets in a checkout line
Selling borrowed seconds back as time

[Pre-Chorus: Daron Malakian]
Turn the dial, divide the sky
Ask the wires to tell us why

[Chorus: Daron Malakian]
Wake up the walls, let the ceiling shake
We built a feast from a plastic lake
Hands in the air while the engines stall
Who signed the order to swallow it all?

[Verse 2: Daron Malakian]
Velvet sirens in the avenue
Paint every warning a pleasant blue
I fed my shadow to the voting machine
It came back polished, quiet, and clean

[Bridge: Daron Malakian]
Softly, softly, count to four
Kick down the arithmetic door
One for the hunger, two for the show
Three for the ones who already know

[Final Chorus: Daron Malakian]
Wake up the walls, let the ceiling shake
We built a feast from a plastic lake
Hands in the air while the engines stall
Who signed the order to swallow it all?

[Outro]
Count the teeth
Kill the light
Carry the silence
Into the night

Inference notes

Each set of four LoRA strengths was generated in one batched pass with a shared random stream. NextLat predictor and XM route-embedding sidecars are training-only and are not used at inference. MiniMax Music 3 may end a song before the configured duration cap.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for terminusresearch/minimax-music3-identity-tournament

Adapter
(35)
this model