MoT

Local AI StudioBETA

Neural AI models running on your graphics card via WebGPU. Private in-browser speech, vision, and text with zero server computation.

๐Ÿค–WebGPU ยท 100% Private ยท Works Offline
Hardware Check
Detecting WebGPUโ€ฆ

Looking for a compatible graphics adapter.

Local Model Storage
Checkingโ€ฆ stored in this browser

0 cached model files. Models are downloaded once and reused offline. Nothing is stored on our servers.

Click to launch

Nothing downloads automatically. Every model shows its size and requirements first.

One-time download

The model is cached in your browser storage and reused on every future visit, even offline.

Runs on your GPU

WebGPU executes the neural network locally. Your files and text never leave your device.

AI Speech-to-Text (Whisper)

Audio ยท Speech Recognition

Transcribe audio files to text 100% locally with OpenAI Whisper. Features 3 selectable model tiers (Whisper Tiny English, Whisper Base Multilingual, and Whisper Small Pro High-Accuracy) supporting Dutch, German, French, Spanish & English with zero network transmissions.

Whisper Model Architecture & Accuracy Fast English ยท ~50 MB
Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
whisper-tiny.en
Download
~50 MB (one-time)
Video memory
~1 GB
Privacy
100% local

AI Background Remover (MODNet)

Vision ยท Image Segmentation

Remove backgrounds from portraits, product photos, and graphics 100% locally in your browser using neural alpha-matting (MODNet). Exports high-resolution transparent PNGs without uploading a single pixel.

Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
modnet
Download
~7 MB (one-time)
Video memory
~1 GB
Privacy
100% local

AI Text Summarizer (DistilBART & BART)

NLP ยท Summarization

Condense long articles, documentation, contracts, or Whisper transcripts into concise executive summaries or bullet points locally using selectable DistilBART CNN (fast) or deep 12-layer BART Large CNN (pro heavy).

Summarizer Model Architecture & Depth Fast Executive ยท ~280 MB
Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
distilbart-cnn-6-6
Download
~280 MB (one-time)
Video memory
~1.5 GB
Privacy
100% local

AI Image Captioning & Alt-Text Generator (ViT-GPT2)

Vision ยท SEO & Accessibility

Analyze photos and illustrations locally using a Vision Encoder-Decoder (ViT + GPT-2) to generate descriptive image captions, WCAG-compliant SEO alt-text attributes, and ready-to-paste HTML <img> snippets.

Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
vit-gpt2-image-captioning
Download
~110 MB (one-time)
Video memory
~1 GB
Privacy
100% local

AI Sentiment & Tone Analyzer (DistilBERT & BERT)

NLP ยท Classification

Evaluate emotional polarity, confidence scores, and sentence-by-sentence tone across customer reviews and support tickets locally using DistilBERT English or granular 5-star BERT Multilingual for Dutch, English, German, and French.

Sentiment Model Architecture & Scope Fast English ยท ~65 MB
Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
distilbert-base-uncased-finetuned-sst-2-english
Download
~65 MB (one-time)
Video memory
~1 GB
Privacy
100% local

AI Local Chat & Code Assistant (Qwen2.5)

LLM ยท Private Generation

Run local Large Language Models (Qwen2.5 0.5B General Assistant & Qwen2.5-Coder Specialist) directly on your GPU with real-time token streaming for 100% private Q&A, coding help, and text rewriting.

AI Model Architecture & Size General Q&A ยท ~460 MB
Local WebGPU AI Model

This model runs entirely on your device. Nothing starts or downloads until you click the button below.

Model
Qwen2.5-0.5B-Instruct
Download
~460 MB (one-time)
Video memory
~1.5 GB
Privacy
100% local
Desktop LLM Engine Recommended: This Large Language Model requires sustained streaming execution. Best experienced on a desktop PC or laptop with WebGPU hardware acceleration.
Private by design: Your audio, images, text, and files are processed exclusively inside your browser. Only open-source files are downloaded once and then cached locally: the model weights from Hugging Face and the ONNX Runtime engine from the jsDelivr CDN. No input data is ever uploaded or logged.
โœจ