RIVI NYX Research Lab

Core Hypothesis

AI character consistency is not only a dialogue problem.

When creators use AI across text, lyrics, vocals, music videos, and visual identity, the question changes: how can one character, voice, or creative identity remain recognizable across multiple generative systems?

中心仮説

AIキャラクターの一貫性は、テキスト対話だけの問題ではない。AIを、会話・歌詞・声・音楽・MV・ビジュアル表現に横断して使うとき、問うべきことは変わる。

RIVI NYX MDRPは、この問題をクロスモーダルな創作アイデンティティの一貫性として捉える。

RIVI NYX MDRP Definition

A creator-facing framework for memory, role, voice, and creative identity.

RIVI NYX MDRP is a prompt-design driven framework for maintaining memory, role, voice, and creative identity across AI characters, lyrics, vocal direction, music generation, MV prompts, and visual identity systems.

It is not proposed as a replacement for academic benchmarks. It is a creator-facing bridge between research concepts and real creative workflows.

Related but distinct concepts: independent naming overlap, conceptual proximity, different origin and application domain.
Area	RIVI NYX MDRP	Academic MDRP
Origin	Creative AI workflow practice and creator-facing prompt design.	Memory-Driven Role-Playing research for persona knowledge utilization in LLMs.
Scope	Characters, lyrics, vocals, music, MV prompts, visual identity, creator QA.	Text role-play, persona memory, evaluation, and enhancement methods.
Known concepts	Anchoring, Recalling, Bounding, Enacting for creative identity continuity.	MREval, MRPrompt, MRBench, Narrative Schema, Magic-If Protocol.
Positioning	Practical bridge between research concepts and creative production workflows.	Academic framework and benchmark direction for LLM persona consistency.

Practical Failure Model

What drift looks like from the creator side

These failure modes describe visible creator-side behavior, not hidden implementation details of any platform.

Persona driftA character gradually sounds like a different person or becomes generic.

Identity driftThe creative identity loses recognizable motifs, values, or emotional shape.

EchoingThe output mirrors the user too directly instead of preserving role logic.

Prompt compressionImportant constraints may be softened, omitted, or over-summarized.

Orchestrator reinterpretationGenerated behavior can appear compressed, redirected, or given added priors.

Memory recall failurePrior context, project history, or recurring motifs fail to reappear.

Boundary collapseThe AI crosses role, tone, safety, or format boundaries that should stay stable.

Vocal persona collapseVoice direction loses emotional range, age, texture, or performance cues.

Lyric structure collapseVerse, chorus, drop, bridge, or narrative arc stops serving its role.

Visual identity driftMotifs, color logic, silhouette, or symbolic constraints shift across outputs.

MV character inconsistencyThe same character appears visually or behaviorally inconsistent across scenes.

MDRP Mechanism

Anchoring / Recalling / Bounding / Enacting

RIVI NYX MDRP treats consistency as an operational design loop: define the core, recall the right context, bound the output, and enact it in the target medium.

1. Anchoring

Character core, values, role, identity phrase, visual identity, vocal identity, and creative intent.

2. Recalling

Memory injection, prior context, project history, recurring motifs, lyric references, and research notes.

3. Bounding

What should remain stable, what should not be mirrored, what counts as drift, and constraints for voice, visuals, and structure.

4. Enacting

Converting memory and boundaries into tone, text, lyric structure, vocal prompt, MV direction, and QA reports.

Research Positioning

What is confirmed, interpreted, and still open

RIVI NYX MDRP may extend existing persona consistency research into creative AI workflows, but it should be treated as a practical interpretation rather than a settled academic claim.

Confirmed Research Facts

Existing studies examine persona drift, long-term memory, role-playing agents, and memory-based evaluation. Academic MDRP proposes MREval / MRPrompt / MRBench.

RIVI NYX Interpretation

Consistency is not solved by persona prompts alone. It involves memory, role, boundary, and enactment.

Open Research Directions

Vocal persona consistency, AI music character consistency, MV identity, and creator-facing evaluation remain open directions.

Creative AI Applications

Where the framework becomes useful

AI Character Design

Reusable identity cores, memory sheets, role boundaries, and tone / voice consistency.

LLM Role-Playing

Reducing echoing, preserving role logic, and reinforcing long-term context.

AI Music Generation

Lyric structure consistency, vocal persona instruction, emotional arc preservation, and chorus / verse role consistency.

MV / Visual Identity

Recurring motifs, color systems, character silhouette constraints, and avoiding identity drift across scenes.

Orchestrator Layer Analysis

Compare user prompt vs generated behavior and document reproducible cases without claiming hidden internals.

Creator QA

Prompt A/B tests, output comparison logs, drift checklists, and creator-facing tips.

Open Research Direction

From text persona to creative identity

Most existing persona consistency research focuses on text-based dialogue. RIVI NYX MDRP extends this question into creative AI workflows: vocal persona, song structure, music video identity, and cross-modal character consistency.

既存研究の多くは、テキスト対話におけるペルソナ一貫性を中心に扱っている。 RIVI NYX MDRPはこの問いを、声・音楽・映像・キャラクター表現を横断する Creative AI Workflowへ拡張する。

Living Source Map

Research Map

This map is rendered from data/research-map.json. It stays editable without a build step.

Loading research map...

Roadmap

Next Steps

Build an MDRP checklist.
Create reusable prompt templates.
Test across AI music models.
Compare outputs across prompt variants.
Define vocal persona consistency criteria.
Maintain research-map.json as a living source map.
Publish Zenn / Note articles.
Create a creator-facing MDRP guide.
Add bilingual JP/EN article pages later.

Articles

Coming Soon

Planned articles are framed as practical research notes for creators.

What is MDRP?

Defines RIVI NYX MDRP and clarifies the academic MDRP naming overlap.

Why Persona Drift Happens

Explains drift as a visible failure of memory, boundary, and enactment.

Prompt Fidelity and Orchestrator Layers

Shows how to compare user prompts against generated behavior.

MDRP for AI Music Creators

Applies MDRP to lyrics, vocal direction, song structure, and emotional arcs.

Vocal Persona Consistency

Explores how vocal tone, age, texture, and emotional range can drift.

MDRP Checklist for Creators

Turns the framework into a repeatable QA checklist for prompts and outputs.

About

Research wording, creator practice, and careful claims

RIVI NYX Research Lab documents practical creative AI observations while keeping research wording conservative. The lab separates confirmed research facts, RIVI NYX interpretation, and open research directions.

本サイトは、創作現場で観測されるAIの一貫性問題を整理するための研究ハブです。学術研究の置き換えではなく、クリエイターが使える設計・記録・検証の言葉へ橋渡しすることを目的とします。