Jun 6 • Tools | Develop Your Skill Set
Character Voice Locking System (CVLS) | My VO solution idea
Had an idea so I sent it to a developer. Wish me luck :) if you are interested in, let me know, no deal has been struck yet.
Character Voice Locking System (CVLS) — A Feature Proposal for Resemble AI
Submitted by: Jonathan Wilson / Studio of Worlds Category:
Product Feature Proposal Date: June 2026
The Problem
AI voice generation has solved quality. It has not solved consistency. Creators working in long-form narrative — animation, serialized storytelling, games, film — need characters who sound the same across hundreds of scenes, across months of production, across the full emotional range of a performance. Currently, achieving this requires rebuilding voice parameters from memory on every session, with no systematic way to lock, save, or reuse a complete character voice profile. The image generation community solved this problem with model sheets and character ICILs (Image Consistency and Identity Lock). Voice generation needs the same solution.
The Proposal: CVLS — Character Voice Locking System
A saved, reusable character voice profile that functions as the audio equivalent of a model sheet.
Layer 1 — Base Voice Profile (The Lock) A permanent character record containing:
Voice ID (selected from Resemble AI library or cloned) Gender Approximate age Ethnicity / cultural background Accent and regional dialect Baseline speaking pace Baseline pitch and tone descriptor Character notes (e.g. "raspy from years of smoke", "formal register", "never raises voice") This profile is saved, named, and recalled instantly. It is the character. It does not drift.
Layer 2 — Emotional Spectrum Mapping A pre-defined set of emotional states mapped to the locked voice, each with adjustable intensity on a 1–10 scale: Neutral (baseline) Joy / Happiness Anger Rage Sadness / Grief Fear Disgust Surprise Tenderness Laughter Crying Whispering Shouting / Yelling Sarcasm Exhaustion Each state is not a binary switch but a dial. "Anger at 3" reads as irritation. "Anger at 9" reads as barely controlled fury. The creator selects the state and the intensity before generating.
Layer 3 — Scripted Generation with Emotional Tagging Within the generation interface, the creator: Selects a saved CVLS profile Writes or pastes the script Tags emotional states and intensities inline, e.g.: [calm:4] "I told you once." [anger:7] "I won't tell you again." [rage:9] "GET OUT." Generates a performance that follows the emotional arc of the script, not just the words.
Why This Matters
Every professional production requires character consistency over time. A character who sounds slightly different in episode 3 than in episode 1 breaks immersion.
A character who cannot modulate between emotional states within a single scene cannot carry a narrative. The current workflow forces creators to reconstruct voice parameters manually, approximate emotional states through prompt language, and accept inconsistency as a cost of production. CVLS eliminates that cost entirely.
Fingers crossed :)
3
16 comments
Character Voice Locking System (CVLS) | My VO solution idea
powered by
skool.com/ai-cinematic-filmmaker-guild-7543
Become the mad scientist. Wield the magic wand. Join the Guild and behold the wonder of AI Filmmaking.
Suggested communities
Powered by