Abdullahu5mani/flowscribe-qwen2.5-0.5b

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:0.5BQuant:BF16Ctx Length:32kPublished:Mar 3, 2026License:mitArchitecture:Transformer Open Weights Warm

Abdullahu5mani/flowscribe-qwen2.5-0.5b is a 500 million parameter instruction-tuned model based on Qwen2.5-0.5B-Instruct, specifically fine-tuned to transform raw speech-to-text transcripts into clean, formatted text. It excels at post-processing messy voice dictation output by removing filler words, correcting grammar, and applying various writing styles such as Professional, Casual, Software_Dev, and Verbatim. This model is optimized for speed and local deployment, making it ideal for applications requiring rapid, style-aware text formatting from speech inputs.

Loading preview...

FlowScribe: Qwen2.5-0.5B Speech Transcript Formatter

Abdullahu5mani/flowscribe-qwen2.5-0.5b is a specialized 500 million parameter language model, fine-tuned from Qwen2.5-0.5B-Instruct, designed to address the common issues with raw speech-to-text outputs. It transforms unformatted, error-prone transcripts (often containing filler words, self-corrections, and missing punctuation) into polished, readable text.

Key Capabilities

  • Multi-Style Formatting: Supports six distinct output styles: Auto (intelligent default), Professional, Casual, Verbatim, Software_Dev, and Enthusiastic.
  • Grammar and Punctuation Correction: Automatically fixes grammatical errors, applies appropriate punctuation, and handles self-corrections.
  • Filler Word Removal: Efficiently strips common filler words like "um" and "uh" to improve clarity.
  • Specialized Formatting: The Software_Dev style, for instance, correctly formats code terms, variable names (e.g., camelCase, snake_case), and technical jargon.
  • Local Deployment Optimized: With approximately 500 million parameters, it prioritizes fast inference and efficient deployment on consumer hardware, including CPU/GPU via a provided Q4_K_M quantized GGUF version.

Good for

  • Post-processing ASR Outputs: Ideal for cleaning up transcripts generated by speech-to-text engines like Whisper.
  • Content Creation: Enhancing dictated articles, reports, or creative writing pieces by applying a desired style.
  • Developer Tools: Formatting spoken code or technical discussions into structured text for documentation or code generation.
  • Accessibility Applications: Improving the readability of voice-dictated notes or communications.
  • Resource-Constrained Environments: Its small size and GGUF quantization make it suitable for local, on-device processing where larger models are impractical.