Hugging FaceModelFritter Nemotron 0.6B Sherpa Onnx Int8danielrosehill/Fritter-Nemotron-0.6B-sherpa-onnx-int8View model on Hugging FaceHugging FaceModelFritter Nemotron 0.6Bdanielrosehill/Fritter-Nemotron-0.6BView model on Hugging FaceHugging FaceModelContextual Sentence Creator GGUFdanielrosehill/Contextual-Sentence-Creator-GGUFGGUF-quantized versions of the Granite-based Term-Context Sentence Generator, optimized for efficient local text generation via llama.cpp.View model on Hugging FaceHugging FaceModelContextual Sentence Creatordanielrosehill/Contextual-Sentence-CreatorA specialized LoRA fine-tune of IBM Granite 3B designed to generate natural English sentences for ASR training data.Read the model cardHugging FaceModelS1 ASR Cleanupdanielrosehill/S1-ASR-CleanupA fine-tuned version of S1-mini designed to transform raw speech-to-text dictation into polished, readable prose.View model on Hugging FaceHugging FaceModelHey Yossi Microwakeworddanielrosehill/Hey-Yossi-microWakeWordA custom wake-word model for the microWakeWord framework, optimized for on-device keyword spotting on resource-constrained hardware.View model on Hugging FaceHugging FaceModelHey Yossi Openwakeworddanielrosehill/Hey-Yossi-openWakeWordA custom wake-word model for openWakeWord designed to detect the phrase 'Hey Yossi' in real-time on CPU-based devices.View model on Hugging FaceHugging FaceModelFritterdanielrosehill/FritterA fine-tuned Automatic Speech Recognition (ASR) model based on NVIDIA Parakeet, optimized for personalized English dictation and technical vocabulary.Read the model cardHugging FaceModelAgent Task Break Detector Modernbertdanielrosehill/Agent-Task-Break-Detector-ModernBERTA 150M-parameter ModernBERT cross-encoder that classifies whether a new user message is an unrelated task that a fresh AI agent session could handle, enabling automatic context resets. It ships as fp32 safetensors and a 151 MB int8 ONNX model for local CPU inference.Read the model cardHugging FaceModelMinilm L12 ASR Paragraph Breaksdanielrosehill/MiniLM-L12-ASR-Paragraph-BreaksA 33.2M-parameter MiniLM-L12 token classifier that inserts paragraph breaks into flat speech-to-text output without altering the transcript text. It ships as a 32 MB int8 ONNX model for on-device use.Read the model cardHugging FaceModelProximate Pseudonymizer 0.6Bdanielrosehill/Proximate-Pseudonymizer-0.6BA fine-tuned 0.6B Qwen3 text-generation model that rewrites text by replacing identifying details with proximate, consistent substitutes, currently a proof of concept that leaves about 9% of identifying details unchanged.Read the model cardHugging FaceModelGemma 3 270M Email Key Line Boldingdanielrosehill/Gemma-3-270M-Email-Key-Line-BoldingA fine-tuned Gemma 3 270M text-generation model that adds Markdown bold to the key requests, purpose, points and facts of a dictated email while leaving the rest of the text unchanged. It is a preliminary release trained mostly on LLM-labelled data, reaching a word-level F1 of 0.62 on a six-email test set.Read the model cardHugging FaceModelGemma 3 270M ASR Paragraph Breaks ONNXdanielrosehill/Gemma-3-270M-ASR-Paragraph-Breaks-ONNXAn ONNX Runtime export of a Gemma 3 270M model fine-tuned to insert paragraph breaks into ASR transcripts, designed to run on-device alongside ONNX speech recognition models. It includes a full-precision and an experimental int8-quantized variant.View model on Hugging FaceHugging FaceModelGemma 3 270M ASR Paragraph Breaksdanielrosehill/Gemma-3-270M-ASR-Paragraph-BreaksA fine-tuned Gemma 3 270M model that adds paragraph breaks to flat speech-to-text output without altering any other characters, designed to run on-device after an ASR pass. It scores 0.78 F1 on held-out synthetic voice memos.Read the model cardHugging FaceModelParakeet TDT 0.6B V2 English Hebrew Codeswitching GGUFdanielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching-GGUFA GGUF-quantized version of the Parakeet-TDT-0.6B-v2 speech recognition model, adapted for English-Hebrew code-switched transcription and designed for use with transcribe.cpp. It is published on Hugging Face under a CC-BY-4.0 license.View model on Hugging FaceHugging FaceModelParakeet TDT 0.6B V2 English Hebrew Codeswitching ONNXdanielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching-ONNXAn ONNX-format export of the Parakeet-TDT-0.6B-v2 speech recognition model, fine-tuned for English-Hebrew code-switching and packaged for use with the onnx-asr runtime. It is published on Hugging Face under a CC-BY-4.0 license.View model on Hugging FaceHugging FaceModelInventory Spec Extractordanielrosehill/inventory-spec-extractorA quantized ONNX token-classification model that extracts structured attributes such as voltage, current, connector type, and serial number from single lines of free-text product specifications. It is fine-tuned from BAAI/bge-small-en-v1.5 and runs on CPU in the browser or on a small server.View model on Hugging FaceHugging FaceModelParakeet TDT 0.6B V2 English Hebrew Codeswitchingdanielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-CodeswitchingA fine-tuned version of NVIDIA's Parakeet-TDT-0.6B-v2 speech recognition model that transcribes English-Hebrew code-switched speech. It is distributed in NeMo, ONNX, and GGUF formats and includes speaker adaptation support.View model on Hugging FaceHugging FaceModelVoice Note Intent Classifierdanielrosehill/Voice-Note-Intent-ClassifierA multi-label intent classifier that reads voice-note transcripts and predicts routing labels such as shopping lists, to-dos, and AI prompts, built on frozen BGE sentence embeddings with per-label logistic regression heads. The v0.1.0 placeholder release reaches a micro F1 of 0.56 on real notes.View model on Hugging FaceHugging FaceModelHebrew Latin Token Classifierdanielrosehill/hebrew-latin-token-classifierA Gradio demo Space for a token classifier that detects Hebrew words written in Latin letters within English text, so text-to-speech systems can pronounce them as Hebrew. It is built on the danielrosehill/hebrew-latin-token-classifier model and published from the Hebrish-ASR repository.View model on Hugging FaceHugging FaceModelJerusalem Poster Detectordanielrosehill/jerusalem-poster-detectorA fine-tuned YOLO11n object detector that locates specific street poster artworks in photographs and identifies which known design it found. It is a proof of concept trained on 13 images, with held-out evaluation showing limited generalisation.View model on Hugging FaceHugging FaceModelWhisper Hebrishdanielrosehill/Whisper-HebrishView model on Hugging FaceHugging FaceModelHerman Poppleberrydanielrosehill/Herman-PoppleberryView model on Hugging FaceHugging FaceModelDinosaur Slothdanielrosehill/Dinosaur-SlothView model on Hugging FaceHugging FaceModelTel Aviv Street Styledanielrosehill/Tel-Aviv-Street-StyleView model on Hugging FaceHugging FaceModelJerusalem Imagesdanielrosehill/Jerusalem-ImagesView model on Hugging Face
For the most up-to-date collection of models, visit my Hugging Face profile