skills.rest
cuba6112
cuba6112COMMUNITY
Explore all skills by this author →
GitHub

unsloth-stt

Fine-tune OpenAI Whisper STT models using Unsloth's optimized LoRA pipeline.

Updated Aug 23, 2026
Search Tags:#fine-tuning#whisper#audio processing#lora#stt#unsloth
One-click install
npx skills add https://github.com/cuba6112/skillfactory --skill unsloth-stt
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: unsloth-stt
Source: https://github.com/cuba6112/skillfactory/tree/main/skills/unsloth-stt
Command: npx skills add https://github.com/cuba6112/skillfactory --skill unsloth-stt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires unsloth, datasets, librosa, transformers, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill addresses the challenge of low transcription accuracy for specialized terminology or accents by enabling efficient fine-tuning of Speech-to-Text (STT) models like Whisper.

Core Features & Use Cases

  • Optimized LoRA Fine-tuning: Leverages Unsloth's pipeline for 1.5x faster training and 50% less memory usage.
  • Specialized Terminology Adaptation: Fine-tune models to accurately transcribe medical, legal, or other domain-specific language.
  • Accent and Dialect Customization: Adapt models to specific regional accents or dialects for improved recognition.
  • Consumer Hardware Training: Enables fine-tuning of large models (e.g., whisper-large-v3) on GPUs with limited VRAM.

Quick Start

Use the unsloth-stt skill to fine-tune the whisper-large-v3 model with your custom audio dataset.

Frequently Asked Questions about unsloth-stt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fine-tune Whisper for specialized terminology and accents?▼

Fine-tune Whisper models using Unsloth's optimized LoRA pipeline to improve transcription accuracy for specialized terminology and regional accents. This adaptation process leverages your custom audio datasets to enhance domain-specific recognition.

Can I fine-tune large Whisper models on consumer hardware with limited VRAM?▼

Yes, you can fine-tune large Whisper models like whisper-large-v3 on consumer hardware with limited VRAM. Unsloth's pipeline reduces memory footprint by 50% and increases training speed by 1.5x, enabling efficient local training.

What audio preprocessing is required before fine-tuning a Speech-to-Text model?▼

Audio preprocessing for Speech-to-Text fine-tuning requires resampling your audio to 16kHz and normalizing the transcripts. These steps ensure your custom dataset matches the input expectations of the Whisper model.

Does Unsloth work with the Hugging Face transformers library for STT model training?▼

Yes, Unsloth works with the transformers library for Speech-to-Text training. It provides an optimized LoRA pipeline that integrates with your datasets and audio processing tools to streamline Whisper fine-tuning.

What is the best way to reduce memory usage when training Speech-to-Text models?▼

The best way to reduce memory usage during Speech-to-Text training is using Unsloth's LoRA pipeline. It achieves a 50% reduction in memory footprint while accelerating training speed by 1.5x for Whisper models.

Related Skills

Explore More Software Engineering Skills→
mxzinkemxzinke
2

stt

Transcribe audio recordings to text via a Whisper-compatible STT backend.

Community
Intermediate
zangjeicyzangjeicy

whisper

Transcribe multilingual audio to text using OpenAI's Whisper model.

Community
Intermediate
HarriesHarries

clip

Connects images and text in a shared embedding space using CLIP.

Community
Intermediate
GarrettRoiGarrettRoi

peft-fine-tuning

Fine-tune large language models with LoRA and QLoRA on limited GPU memory.

Community
Advanced