My text-only sentence banks and paragraph sets for reading aloud and ASR training. No audio.
Daniel Rosehill PRO
danielrosehill
AI & ML interests
Speech to text (ASR), voice workflows, MCP, AI agents, orchestration, automation.
Recent Activity
updated a dataset less than a minute ago
Hebrish/Singing-Sentences updated a dataset less than a minute ago
Hebrish/Repeated-Word-Sentences updated a dataset less than a minute ago
Hebrish/Brand-Name-SentencesOrganizations
Text-only sentence datasets by others, candidates for reading lists and recording prompts. No audio.
-
agentlans/high-quality-english-sentences
Viewer • Updated • 1.71M • 893 • 40 -
nineninesix/harvard-sentences-tts-benchmark
Viewer • Updated • 720 • 37 • 3 -
tomron87/hebrew-wikipedia-sentences-corpus
Viewer • Updated • 11M • 46 -
sentence-transformers/parallel-sentences-tatoeba
Viewer • Updated • 8.35M • 1.92k
Open sourced audio datasets for STT/ASR. All recordings by me (Daniel Rosehill) unless otherwise accredited.
Personal ASR fine-tune for tech domain vocabulary recognition and integrated formatting tasks
Early stage proof of concept models that require more training data, training, or rearchitecting.
Datasets packaging (third-party) open access datasets for AI/ML use
Base fine tune and variants of Parakeet V2 English-Hebrew codeswitching finetune. Website: hebrish.app
-
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k -
Hebrish/Jachnun-GGUF
Automatic Speech Recognition • 0.6B • Updated • 222 -
Hebrish/Jachnun-ONNX
Automatic Speech Recognition • Updated • 36 -
Hebrish/Jachnun-sherpa-onnx-int8
Automatic Speech Recognition • Updated
Candidates for a fine tuning idea for audio to text with deterministic post processing edits applied in one forward pass
-
nvidia/canary-1b
Automatic Speech Recognition • Updated • 3.04k • 463 -
Qwen/Qwen3-ASR-1.7B-hf
Automatic Speech Recognition • 2B • Updated • 336k • 162 -
nvidia/nemotron-3.5-asr-streaming-0.6b
Automatic Speech Recognition • 0.6B • Updated • 573k • • 1.26k -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k
A few models to try in Frigate for presence detection of pigeons.
-
meta-llama/Llama-3.2-1B-Instruct
Text Generation • 1B • Updated • 1.44M • • 1.96k -
Qwen/Qwen3-0.6B
Text Generation • 0.8B • Updated • 21M • • 1.88k -
empero-ai/Qwen3.8-2B-Distill-GGUF
Text Generation • 2B • Updated • 511k • 181 -
danielrosehill/Contextual-Sentence-Creator
Text Generation • 4B • Updated • 373
-
st1992/bert-restore-punctuation
Token Classification • Updated • 15 -
superwhisper/s1-mini
Text Generation • 0.8B • Updated • 1.91k • 386 -
oliverguhr/fullstop-punctuation-multilang-large
Token Classification • 0.6B • Updated • 350k • • 183 -
superwhisper/s1-mini-GGUF
Text Generation • 0.8B • Updated • 239k • 46
Models that apply a text-to-text transformation as a second pass on speech transcribed using ASR.
-
zai-org/GLM-ASR-Nano-2512
Automatic Speech Recognition • 2B • Updated • 19.8k • 388 -
nvidia/canary-qwen-2.5b
Automatic Speech Recognition • 3B • Updated • 22.2k • 474 -
microsoft/Phi-4-multimodal-instruct
Automatic Speech Recognition • 6B • Updated • 76.7k • 1.63k -
facebook/omniASR-LLM-7B
Automatic Speech Recognition • Updated • 47
The two official variants of Voxtral (audio multimodal model) released by Mistral in July 2025
-
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
ibm-granite/granite-speech-3.3-8b
Automatic Speech Recognition • 9B • Updated • 5.63k • 171 -
facebook/seamless-m4t-v2-large
Automatic Speech Recognition • 2B • Updated • 299k • 1.04k -
facebook/wav2vec2-base-960h
Automatic Speech Recognition • 94.4M • Updated • 1.74M • 417
-
futo-org/acft-whisper-tiny
Automatic Speech Recognition • 57.7M • Updated • 54 • 2 -
futo-org/acft-whisper-small.en
Automatic Speech Recognition • 0.3B • Updated • 20 • 2 -
futo-org/acft-whisper-base.en
Automatic Speech Recognition • 99.1M • Updated • 17 • 2 -
futo-org/acft-whisper-tiny.en
Automatic Speech Recognition • 57.7M • Updated • 15 • 1
Models that work in unison with core STT models in voice workflows
-
pyannote/voice-activity-detection
Automatic Speech Recognition • Updated • 303k • 243 -
pyannote/speaker-diarization-3.1
Automatic Speech Recognition • Updated • 6.78M • 4.99k -
pyannote/overlapped-speech-detection
Automatic Speech Recognition • Updated • 1.97k • 66 -
pipecat-ai/smart-turn-v3
Voice Activity Detection • Updated • 217
-
unsloth/whisper-large-v3
Automatic Speech Recognition • 2B • Updated • 2.73k • 20 -
unsloth/whisper-small
Automatic Speech Recognition • 0.2B • Updated • 2.58k • 7 -
unsloth/CrisperWhisper
Automatic Speech Recognition • 2B • Updated • 809 • 17 -
unsloth/whisper-large-v3-turbo
Automatic Speech Recognition • 0.8B • Updated • 3.04k • 13
-
Baby Noise Cancellation Demo
👶AI-powered baby noise removal demo with STT comparison
-
Nano Banana Sketch Cleanup
🎨AI sketch cleanup with before/after comparisons
-
Email Key-Line Bolding
✉Bold or highlight the key lines of an email (Gemma 270M)
-
ASR Paragraph Breaks
🎙Add paragraph breaks to flat ASR text (Gemma 270M)
-
Pharmacology Graph Explorer
💊3Explore predicted drug‑target and disease links
-
Medical Diagnosis
📉77Classify symptoms into possible medical conditions
-
MediAI Medical AI Agent
🚀29AI-Powered Diagnosis & Treatment Assistant
-
Lisdexamfetamine Split Dose Modeller
🚀Model split-dose protocols for lisdexamfetamine/Vyvanse
Spaces and models that may have applications in open source intelligence (OSINT)
AI use-cases and appliations that I found interesting (a repo for myself to explore!)
LLMs optimised for instruction following rather than conversational use - quants and original models
-
meta-llama/Llama-3.1-8B-Instruct
Text Generation • 8B • Updated • 4.35M • • 8.84k -
Qwen/Qwen3-Omni-30B-A3B-Instruct
Any-to-Any • 35B • Updated • 435k • 1.05k -
moonshotai/Kimi-K2-Instruct-0905
Text Generation • 1T • Updated • 9.38k • • 790 -
mistralai/Mistral-7B-Instruct-v0.3
7B • Updated • 1.73M • 3.83k
LLMs optimised for running "on-device" (specifically, for this collection, on smartphones with standard inference capabilities)
LLMs (mostly quants) that are small enough to run locally (on my hardware). My go-tos.
-
Qwen/Qwen3-VL-8B-Instruct
Image-Text-to-Text • 9B • Updated • 7.11M • • 1.25k -
zai-org/GLM-4.6
Text Generation • 357B • Updated • 27.6k • • 1.25k -
Qwen/Qwen3-VL-8B-Thinking
Image-Text-to-Text • 9B • Updated • 133k • • 234 -
meta-llama/Llama-3.1-8B-Instruct
Text Generation • 8B • Updated • 4.35M • • 8.84k
AI agent network configs and projects
Projects related to Israel
Datasets for STT training, classification fine-tuning - especially voice notes
Implementations of apps for simple text reformatting tasks
Spaces to do with Hebrew language AI
Language models supporting TTS (predominantly the text generation task). These are almost exclusively fine-tunes of Gemma, Mistral, or the Qwen models
Text-only sentences and short paragraphs for reading aloud, in English and Hebrew: mine and third party. No audio.
The major public speech and audio datasets, kept for reference in case they are used in training.
-
AtomicChat/DeepSeek-V4.1-Flash-1017-Uncensored-GSQ-RCO-GGUF
Text Generation • 552B • Updated • 130k • 386 -
SC117/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF
Image-Text-to-Text • 177B • Updated • 4.21M • 778 -
audnai/penclaw-Kimi-K3.1-pro-unrestricted
Text Generation • 2.8T • Updated • 157 • 180 -
huihui-ai/Huihui-GLM-5.3-Flash-abliterated-GGUF
Image-Text-to-Text • 321B • Updated • 41.8k • 372
Models that write natural sentences around a given term or context, e.g. read-aloud prompts for recording ASR training data.
Models I'm using / considering using with agentic harnesses - CLIs
-
MiniMaxAI/MiniMax-M3.1-Flash
Image-Text-to-Text • 220B • Updated • 51.2k • • 563 -
deepseek-ai/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 296k • • 5.41k -
Qwen/Qwen3.8-27B
Image-Text-to-Text • 28B • Updated • 7.67M • • 19.4k -
unsloth/flint-GGUF
Text Generation • 40B • Updated • 304k • 368
Single task model for paragraph inference intended as a deterministic single post-processing pass in ASR pipelines.
-
Paragraph Inference for ASR
📝What paragraph inference is, and where it fits in ASR
-
danielrosehill/MiniLM-L12-ASR-Paragraph-Breaks
Token Classification • 33.2M • Updated • 53 • 1 -
danielrosehill/Gemma-3-270M-ASR-Paragraph-Breaks
Text Generation • 0.3B • Updated • 711 -
danielrosehill/Gemma-3-270M-ASR-Paragraph-Breaks-ONNX
Text Generation • Updated • 406
-
verbit/hebrew_punctuation
Updated • 155 • 3 -
MaorB/embeddinggemma-2-asr-hebrew-english-streaming
Automatic Speech Recognition • 99M • Updated • 115 • 3 -
imvladikon/wav2vec2-large-xlsr-53-hebrew
Automatic Speech Recognition • 0.3B • Updated • 6.49k • 8 -
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k
Datasets intended for (multimodal) training of models with understanding of real-life context of Israel - esp. settings and places
Datasets of evaluation runs
Datasets and fine tunes (mine) focused on English (primary) with Hebrew code-switching. Targeted use-case: improved ASR accuracy among olim
-
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k -
danielrosehill/English-Hebrew-Codeswitching-Sentences
Updated • 18.1k • 1 -
danielrosehill/Whisper-Hebrish
Automatic Speech Recognition • 0.8B • Updated • 41 -
danielrosehill/Read-Aloud-Voice-Sentences
Updated • 19.8k
Token classification models. Use-case: Linguistic identification for code switching in TTS script normalization.
Small ASR models that aren't Whisper (or derivs). On device inference for smartphones etc
-
netease-youdao/Confucius4-R2T2
Automatic Speech Recognition • 2B • Updated • 61k • 592 -
oruk/orukeet
Automatic Speech Recognition • 0.6B • Updated • 33.9k • 107 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
Qwen/Qwen3-ASR-1.7B
Automatic Speech Recognition • 2B • Updated • 1.74M • • 1.23k
Models used for on device speech to text with the OnePlus CPH2493.
-
nvidia/parakeet-tdt-1.1b
Automatic Speech Recognition • Updated • 7k • 142 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
nvidia/nemotron-3.5-asr-streaming-0.6b
Automatic Speech Recognition • 0.6B • Updated • 573k • • 1.26k -
handy-computer/parakeet-tdt-1.1b-gguf
Automatic Speech Recognition • 1B • Updated • 24k • 1
A collection intended as a first entry store for small models, especially component models, which perform modular tightly bounded tasks.
Community derivatives from Deepseek V4 and V4.1
-
deepseek-ai/DeepSeek-V4.1-Flash
Image-Text-to-Text • 763B • Updated • 1.84M • • 4.9k -
dealignai/DeepSeek-V4.1-Flash-UNCENSORED-FP8
Image-Text-to-Text • 763B • Updated • 23.1k • 834 -
deepseek-ai/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 296k • • 5.41k -
unsloth/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 1.56k • 11
Speech to text models that I've had the best experience with on my current hardware (One Plus Nord 3 ). Currently this is mostly ASR models.
ASR and multimodal models supporting Hebrew to English through audio and text token processing.
-
imvladikon/wav2vec2-large-xlsr-53-hebrew
Automatic Speech Recognition • 0.3B • Updated • 6.49k • 8 -
Mizurodp/wav2vec2-large-xls-r-300m-hebrew-colab
Automatic Speech Recognition • Updated • 64 • 1 -
imvladikon/wav2vec2-xls-r-300m-lm-hebrew
Automatic Speech Recognition • 0.3B • Updated • 557 • 4 -
imvladikon/wav2vec2-xls-r-1b-hebrew
Automatic Speech Recognition • 1.0B • Updated • 481 • 2
Quick collection of models I'm evaluating for cogeneration. Minimum inclusion criteria includes tool usage and MCP. Looking for something fast and goo
Open source models with audio understanding. Tracking mostly vendor releases in the audio and text to text subclassification of multimodal.
Useful datasets and models for ASR projects and fine tuning
Useful models for demonstrating what early and pre-Transformer LLMs looked and functioned like
Experiments in AI and prompt engineering
-
danielrosehill/Podcast-ASR-Evaluation
Viewer • Updated • 27 • 20 -
danielrosehill/Long-Prompt-Experiment
Viewer • Updated • 1 • 110 -
Podcast ASR Evaluation
🎙ASR benchmark comparing local and cloud models
-
LLM Long Output Experiment (Code Generation)
📈1Evaluating max single output length of code gen LLMs
Collection of models gathered together for my wife who is an architect (of buildings!)
Whisper fine-tuned on 1 hour of my voice
-
nyralabs/CrisperWhisper
Automatic Speech Recognition • 2B • Updated • 14.2k • 346 -
openai/whisper-base
Automatic Speech Recognition • 72.6M • Updated • 4.51M • 313 -
openai/whisper-base.en
Automatic Speech Recognition • 72.6M • Updated • 210k • 61 -
onnx-community/whisper-base_timestamped
Automatic Speech Recognition • Updated • 75.1k • 32
Just for fun projects for converting conventional text into Shakespearean English
Experimentary workflows (mostly) that aim to rethink how contextual data can be gathered to achieve personalised inference in AI systems
-
google/embeddinggemma-300m
Sentence Similarity • 0.3B • Updated • 1.18M • • 2.1k -
nvidia/omni-embed-nemotron-3b
Sentence Similarity • 5B • Updated • 1.98k • 134 -
Qwen/Qwen3-Embedding-0.6B
Feature Extraction • 0.6B • Updated • 9.23M • • 1.34k -
google/embeddinggemma-2
Feature Extraction • 0.7B • Updated • 507k • 3.2k
-
tencent/HunyuanImage-3.0
Text-to-Image • 83B • Updated • 1.23k • • 1.16k -
black-forest-labs/FLUX.1-schnell
Text-to-Image • 12B • Updated • 277k • • 6.65k -
stabilityai/stable-diffusion-xl-base-1.0
Text-to-Image • 3B • Updated • 4.63M • • 8.45k -
Qwen/Qwen-Image
Text-to-Image • 20B • Updated • 330k • • 2.69k
Quantized versions of image generation models
-
danielrosehill/Shakespearean-Text-Transformation-Prompts
Viewer • Updated • 1 • 48 -
danielrosehill/Speech-To-Text-System-Prompts-2
Viewer • Updated • 2 • 103 • 1 -
System Prompt Reformatter
📚Reformats system prompts in the 2nd person and other edits
-
BLUF Email Formatter
📧Format emails with clear subject lines and summaries
Projects related to the city of Jerusalem
ISO standard related projects
Datasets and projects related to sustainability esp impact investing
Collections of system prompts for assistants, agents, and workflows
Version controlled refactors for data analysis of the Global Value Factor Database (GVFD) by the International Foundation for Valuing Impacts (IFVI).
-
modularai/Llama-3.1-8B-Instruct-GGUF
Text Generation • 8B • Updated • 7.71k • 17 -
MaziyarPanahi/WizardLM-2-7B-GGUF
Text Generation • 7B • Updated • 23.5k • 83 -
MaziyarPanahi/mathstral-7B-v0.1-GGUF
Text Generation • 7B • Updated • 22.8k • 8 -
MaziyarPanahi/phi-4-GGUF
Text Generation • 15B • Updated • 23.2k • 10
Datasets in Hebrew
Collection of ASR models for customized STT model training
-
nvidia/parakeet-tdt-0.6b-v2
Automatic Speech Recognition • Updated • 175k • 1.59k -
ibm-granite/granite-speech-3.3-8b
Automatic Speech Recognition • 9B • Updated • 5.63k • 171 -
nvidia/canary-qwen-2.5b
Automatic Speech Recognition • 3B • Updated • 22.2k • 474 -
facebook/omniASR-W2V-1B
Automatic Speech Recognition • Updated • 6
My text-only sentence banks and paragraph sets for reading aloud and ASR training. No audio.
Text-only sentences and short paragraphs for reading aloud, in English and Hebrew: mine and third party. No audio.
Text-only sentence datasets by others, candidates for reading lists and recording prompts. No audio.
-
agentlans/high-quality-english-sentences
Viewer • Updated • 1.71M • 893 • 40 -
nineninesix/harvard-sentences-tts-benchmark
Viewer • Updated • 720 • 37 • 3 -
tomron87/hebrew-wikipedia-sentences-corpus
Viewer • Updated • 11M • 46 -
sentence-transformers/parallel-sentences-tatoeba
Viewer • Updated • 8.35M • 1.92k
The major public speech and audio datasets, kept for reference in case they are used in training.
Open sourced audio datasets for STT/ASR. All recordings by me (Daniel Rosehill) unless otherwise accredited.
-
AtomicChat/DeepSeek-V4.1-Flash-1017-Uncensored-GSQ-RCO-GGUF
Text Generation • 552B • Updated • 130k • 386 -
SC117/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF
Image-Text-to-Text • 177B • Updated • 4.21M • 778 -
audnai/penclaw-Kimi-K3.1-pro-unrestricted
Text Generation • 2.8T • Updated • 157 • 180 -
huihui-ai/Huihui-GLM-5.3-Flash-abliterated-GGUF
Image-Text-to-Text • 321B • Updated • 41.8k • 372
Models that write natural sentences around a given term or context, e.g. read-aloud prompts for recording ASR training data.
Models I'm using / considering using with agentic harnesses - CLIs
-
MiniMaxAI/MiniMax-M3.1-Flash
Image-Text-to-Text • 220B • Updated • 51.2k • • 563 -
deepseek-ai/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 296k • • 5.41k -
Qwen/Qwen3.8-27B
Image-Text-to-Text • 28B • Updated • 7.67M • • 19.4k -
unsloth/flint-GGUF
Text Generation • 40B • Updated • 304k • 368
Personal ASR fine-tune for tech domain vocabulary recognition and integrated formatting tasks
Single task model for paragraph inference intended as a deterministic single post-processing pass in ASR pipelines.
-
Paragraph Inference for ASR
📝What paragraph inference is, and where it fits in ASR
-
danielrosehill/MiniLM-L12-ASR-Paragraph-Breaks
Token Classification • 33.2M • Updated • 53 • 1 -
danielrosehill/Gemma-3-270M-ASR-Paragraph-Breaks
Text Generation • 0.3B • Updated • 711 -
danielrosehill/Gemma-3-270M-ASR-Paragraph-Breaks-ONNX
Text Generation • Updated • 406
Early stage proof of concept models that require more training data, training, or rearchitecting.
-
verbit/hebrew_punctuation
Updated • 155 • 3 -
MaorB/embeddinggemma-2-asr-hebrew-english-streaming
Automatic Speech Recognition • 99M • Updated • 115 • 3 -
imvladikon/wav2vec2-large-xlsr-53-hebrew
Automatic Speech Recognition • 0.3B • Updated • 6.49k • 8 -
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k
Datasets intended for (multimodal) training of models with understanding of real-life context of Israel - esp. settings and places
Datasets packaging (third-party) open access datasets for AI/ML use
Datasets of evaluation runs
Base fine tune and variants of Parakeet V2 English-Hebrew codeswitching finetune. Website: hebrish.app
-
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k -
Hebrish/Jachnun-GGUF
Automatic Speech Recognition • 0.6B • Updated • 222 -
Hebrish/Jachnun-ONNX
Automatic Speech Recognition • Updated • 36 -
Hebrish/Jachnun-sherpa-onnx-int8
Automatic Speech Recognition • Updated
Datasets and fine tunes (mine) focused on English (primary) with Hebrew code-switching. Targeted use-case: improved ASR accuracy among olim
-
danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching
Automatic Speech Recognition • 0.6B • Updated • 3.31k -
danielrosehill/English-Hebrew-Codeswitching-Sentences
Updated • 18.1k • 1 -
danielrosehill/Whisper-Hebrish
Automatic Speech Recognition • 0.8B • Updated • 41 -
danielrosehill/Read-Aloud-Voice-Sentences
Updated • 19.8k
Candidates for a fine tuning idea for audio to text with deterministic post processing edits applied in one forward pass
-
nvidia/canary-1b
Automatic Speech Recognition • Updated • 3.04k • 463 -
Qwen/Qwen3-ASR-1.7B-hf
Automatic Speech Recognition • 2B • Updated • 336k • 162 -
nvidia/nemotron-3.5-asr-streaming-0.6b
Automatic Speech Recognition • 0.6B • Updated • 573k • • 1.26k -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k
Token classification models. Use-case: Linguistic identification for code switching in TTS script normalization.
A few models to try in Frigate for presence detection of pigeons.
Small ASR models that aren't Whisper (or derivs). On device inference for smartphones etc
-
netease-youdao/Confucius4-R2T2
Automatic Speech Recognition • 2B • Updated • 61k • 592 -
oruk/orukeet
Automatic Speech Recognition • 0.6B • Updated • 33.9k • 107 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
Qwen/Qwen3-ASR-1.7B
Automatic Speech Recognition • 2B • Updated • 1.74M • • 1.23k
-
meta-llama/Llama-3.2-1B-Instruct
Text Generation • 1B • Updated • 1.44M • • 1.96k -
Qwen/Qwen3-0.6B
Text Generation • 0.8B • Updated • 21M • • 1.88k -
empero-ai/Qwen3.8-2B-Distill-GGUF
Text Generation • 2B • Updated • 511k • 181 -
danielrosehill/Contextual-Sentence-Creator
Text Generation • 4B • Updated • 373
-
st1992/bert-restore-punctuation
Token Classification • Updated • 15 -
superwhisper/s1-mini
Text Generation • 0.8B • Updated • 1.91k • 386 -
oliverguhr/fullstop-punctuation-multilang-large
Token Classification • 0.6B • Updated • 350k • • 183 -
superwhisper/s1-mini-GGUF
Text Generation • 0.8B • Updated • 239k • 46
Models used for on device speech to text with the OnePlus CPH2493.
-
nvidia/parakeet-tdt-1.1b
Automatic Speech Recognition • Updated • 7k • 142 -
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
nvidia/nemotron-3.5-asr-streaming-0.6b
Automatic Speech Recognition • 0.6B • Updated • 573k • • 1.26k -
handy-computer/parakeet-tdt-1.1b-gguf
Automatic Speech Recognition • 1B • Updated • 24k • 1
A collection intended as a first entry store for small models, especially component models, which perform modular tightly bounded tasks.
Community derivatives from Deepseek V4 and V4.1
-
deepseek-ai/DeepSeek-V4.1-Flash
Image-Text-to-Text • 763B • Updated • 1.84M • • 4.9k -
dealignai/DeepSeek-V4.1-Flash-UNCENSORED-FP8
Image-Text-to-Text • 763B • Updated • 23.1k • 834 -
deepseek-ai/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 296k • • 5.41k -
unsloth/DeepSeek-V4.1-Flash-1017
Image-Text-to-Text • 763B • Updated • 1.56k • 11
Models that apply a text-to-text transformation as a second pass on speech transcribed using ASR.
Speech to text models that I've had the best experience with on my current hardware (One Plus Nord 3 ). Currently this is mostly ASR models.
ASR and multimodal models supporting Hebrew to English through audio and text token processing.
-
imvladikon/wav2vec2-large-xlsr-53-hebrew
Automatic Speech Recognition • 0.3B • Updated • 6.49k • 8 -
Mizurodp/wav2vec2-large-xls-r-300m-hebrew-colab
Automatic Speech Recognition • Updated • 64 • 1 -
imvladikon/wav2vec2-xls-r-300m-lm-hebrew
Automatic Speech Recognition • 0.3B • Updated • 557 • 4 -
imvladikon/wav2vec2-xls-r-1b-hebrew
Automatic Speech Recognition • 1.0B • Updated • 481 • 2
Quick collection of models I'm evaluating for cogeneration. Minimum inclusion criteria includes tool usage and MCP. Looking for something fast and goo
-
zai-org/GLM-ASR-Nano-2512
Automatic Speech Recognition • 2B • Updated • 19.8k • 388 -
nvidia/canary-qwen-2.5b
Automatic Speech Recognition • 3B • Updated • 22.2k • 474 -
microsoft/Phi-4-multimodal-instruct
Automatic Speech Recognition • 6B • Updated • 76.7k • 1.63k -
facebook/omniASR-LLM-7B
Automatic Speech Recognition • Updated • 47
The two official variants of Voxtral (audio multimodal model) released by Mistral in July 2025
Open source models with audio understanding. Tracking mostly vendor releases in the audio and text to text subclassification of multimodal.
-
nvidia/parakeet-tdt-0.6b-v3
Automatic Speech Recognition • 0.6B • Updated • 548k • 1.27k -
ibm-granite/granite-speech-3.3-8b
Automatic Speech Recognition • 9B • Updated • 5.63k • 171 -
facebook/seamless-m4t-v2-large
Automatic Speech Recognition • 2B • Updated • 299k • 1.04k -
facebook/wav2vec2-base-960h
Automatic Speech Recognition • 94.4M • Updated • 1.74M • 417
Useful datasets and models for ASR projects and fine tuning
Useful models for demonstrating what early and pre-Transformer LLMs looked and functioned like
-
futo-org/acft-whisper-tiny
Automatic Speech Recognition • 57.7M • Updated • 54 • 2 -
futo-org/acft-whisper-small.en
Automatic Speech Recognition • 0.3B • Updated • 20 • 2 -
futo-org/acft-whisper-base.en
Automatic Speech Recognition • 99.1M • Updated • 17 • 2 -
futo-org/acft-whisper-tiny.en
Automatic Speech Recognition • 57.7M • Updated • 15 • 1
Experiments in AI and prompt engineering
-
danielrosehill/Podcast-ASR-Evaluation
Viewer • Updated • 27 • 20 -
danielrosehill/Long-Prompt-Experiment
Viewer • Updated • 1 • 110 -
Podcast ASR Evaluation
🎙ASR benchmark comparing local and cloud models
-
LLM Long Output Experiment (Code Generation)
📈1Evaluating max single output length of code gen LLMs
Models that work in unison with core STT models in voice workflows
-
pyannote/voice-activity-detection
Automatic Speech Recognition • Updated • 303k • 243 -
pyannote/speaker-diarization-3.1
Automatic Speech Recognition • Updated • 6.78M • 4.99k -
pyannote/overlapped-speech-detection
Automatic Speech Recognition • Updated • 1.97k • 66 -
pipecat-ai/smart-turn-v3
Voice Activity Detection • Updated • 217
Collection of models gathered together for my wife who is an architect (of buildings!)
-
unsloth/whisper-large-v3
Automatic Speech Recognition • 2B • Updated • 2.73k • 20 -
unsloth/whisper-small
Automatic Speech Recognition • 0.2B • Updated • 2.58k • 7 -
unsloth/CrisperWhisper
Automatic Speech Recognition • 2B • Updated • 809 • 17 -
unsloth/whisper-large-v3-turbo
Automatic Speech Recognition • 0.8B • Updated • 3.04k • 13
Whisper fine-tuned on 1 hour of my voice
-
nyralabs/CrisperWhisper
Automatic Speech Recognition • 2B • Updated • 14.2k • 346 -
openai/whisper-base
Automatic Speech Recognition • 72.6M • Updated • 4.51M • 313 -
openai/whisper-base.en
Automatic Speech Recognition • 72.6M • Updated • 210k • 61 -
onnx-community/whisper-base_timestamped
Automatic Speech Recognition • Updated • 75.1k • 32
-
Baby Noise Cancellation Demo
👶AI-powered baby noise removal demo with STT comparison
-
Nano Banana Sketch Cleanup
🎨AI sketch cleanup with before/after comparisons
-
Email Key-Line Bolding
✉Bold or highlight the key lines of an email (Gemma 270M)
-
ASR Paragraph Breaks
🎙Add paragraph breaks to flat ASR text (Gemma 270M)
Just for fun projects for converting conventional text into Shakespearean English
Experimentary workflows (mostly) that aim to rethink how contextual data can be gathered to achieve personalised inference in AI systems
-
Pharmacology Graph Explorer
💊3Explore predicted drug‑target and disease links
-
Medical Diagnosis
📉77Classify symptoms into possible medical conditions
-
MediAI Medical AI Agent
🚀29AI-Powered Diagnosis & Treatment Assistant
-
Lisdexamfetamine Split Dose Modeller
🚀Model split-dose protocols for lisdexamfetamine/Vyvanse
Spaces and models that may have applications in open source intelligence (OSINT)
AI use-cases and appliations that I found interesting (a repo for myself to explore!)
-
google/embeddinggemma-300m
Sentence Similarity • 0.3B • Updated • 1.18M • • 2.1k -
nvidia/omni-embed-nemotron-3b
Sentence Similarity • 5B • Updated • 1.98k • 134 -
Qwen/Qwen3-Embedding-0.6B
Feature Extraction • 0.6B • Updated • 9.23M • • 1.34k -
google/embeddinggemma-2
Feature Extraction • 0.7B • Updated • 507k • 3.2k
LLMs optimised for instruction following rather than conversational use - quants and original models
-
meta-llama/Llama-3.1-8B-Instruct
Text Generation • 8B • Updated • 4.35M • • 8.84k -
Qwen/Qwen3-Omni-30B-A3B-Instruct
Any-to-Any • 35B • Updated • 435k • 1.05k -
moonshotai/Kimi-K2-Instruct-0905
Text Generation • 1T • Updated • 9.38k • • 790 -
mistralai/Mistral-7B-Instruct-v0.3
7B • Updated • 1.73M • 3.83k
-
tencent/HunyuanImage-3.0
Text-to-Image • 83B • Updated • 1.23k • • 1.16k -
black-forest-labs/FLUX.1-schnell
Text-to-Image • 12B • Updated • 277k • • 6.65k -
stabilityai/stable-diffusion-xl-base-1.0
Text-to-Image • 3B • Updated • 4.63M • • 8.45k -
Qwen/Qwen-Image
Text-to-Image • 20B • Updated • 330k • • 2.69k
Quantized versions of image generation models
LLMs optimised for running "on-device" (specifically, for this collection, on smartphones with standard inference capabilities)
LLMs (mostly quants) that are small enough to run locally (on my hardware). My go-tos.
-
Qwen/Qwen3-VL-8B-Instruct
Image-Text-to-Text • 9B • Updated • 7.11M • • 1.25k -
zai-org/GLM-4.6
Text Generation • 357B • Updated • 27.6k • • 1.25k -
Qwen/Qwen3-VL-8B-Thinking
Image-Text-to-Text • 9B • Updated • 133k • • 234 -
meta-llama/Llama-3.1-8B-Instruct
Text Generation • 8B • Updated • 4.35M • • 8.84k
-
danielrosehill/Shakespearean-Text-Transformation-Prompts
Viewer • Updated • 1 • 48 -
danielrosehill/Speech-To-Text-System-Prompts-2
Viewer • Updated • 2 • 103 • 1 -
System Prompt Reformatter
📚Reformats system prompts in the 2nd person and other edits
-
BLUF Email Formatter
📧Format emails with clear subject lines and summaries
AI agent network configs and projects
Projects related to the city of Jerusalem
Projects related to Israel
ISO standard related projects
Datasets and projects related to sustainability esp impact investing
Datasets for STT training, classification fine-tuning - especially voice notes
Collections of system prompts for assistants, agents, and workflows
Version controlled refactors for data analysis of the Global Value Factor Database (GVFD) by the International Foundation for Valuing Impacts (IFVI).
Implementations of apps for simple text reformatting tasks
-
modularai/Llama-3.1-8B-Instruct-GGUF
Text Generation • 8B • Updated • 7.71k • 17 -
MaziyarPanahi/WizardLM-2-7B-GGUF
Text Generation • 7B • Updated • 23.5k • 83 -
MaziyarPanahi/mathstral-7B-v0.1-GGUF
Text Generation • 7B • Updated • 22.8k • 8 -
MaziyarPanahi/phi-4-GGUF
Text Generation • 15B • Updated • 23.2k • 10
Spaces to do with Hebrew language AI
Datasets in Hebrew
Collection of ASR models for customized STT model training
-
nvidia/parakeet-tdt-0.6b-v2
Automatic Speech Recognition • Updated • 175k • 1.59k -
ibm-granite/granite-speech-3.3-8b
Automatic Speech Recognition • 9B • Updated • 5.63k • 171 -
nvidia/canary-qwen-2.5b
Automatic Speech Recognition • 3B • Updated • 22.2k • 474 -
facebook/omniASR-W2V-1B
Automatic Speech Recognition • Updated • 6
Language models supporting TTS (predominantly the text generation task). These are almost exclusively fine-tunes of Gemma, Mistral, or the Qwen models