PulseAugur
EN
LIVE 19:12:48

Developer blends LLM with n-gram for personalized voice cloning

By PulseAugur Editorial · [1 sources] ·

A developer experimented with voice cloning by combining a small language model (SmolLM2) with a token-level n-gram trained on their own published writings. The goal was to create a chatbot that sounded like the developer, but the results were mixed. While the n-gram added a personal linguistic style, the language model struggled to grasp the underlying concepts, leading to output that was grammatically correct but conceptually empty or paranoid. AI

IMPACT This experiment explores novel methods for personalizing LLM output by combining statistical n-grams with generative models, potentially influencing future approaches to custom AI voices.

RANK_REASON The cluster describes an experiment combining existing models and techniques to achieve a specific outcome, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Developer blends LLM with n-gram for personalized voice cloning

How we ranked this

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster describes an experiment combining existing models and techniques to achieve a specific outcome, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, other
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
123 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Alex Towell ·

    I Tried to Voice-Clone Myself in 90 MB

    <p>The experiment: a tiny SmolLM2 running in your browser. A token-level n-gram trained on every word I have published. Mix the two distributions in probability space at every generation step. Sample from the mix.</p> <p>You can try it at <a href="https://metafunctor.com/ask/" re…

RELATED ENTITIES

RELATED TOPICS