# Classify Text Against Labels You Invent, With No Training

> Give a model any set of categories and it sorts your text into them — no training data, no fine-tuning, no API key.

Zero-shot classification sorts text into categories you invent at runtime, with no training data. It works by reframing classification as natural language inference: your text becomes a premise and each label is turned into a hypothesis whose entailment probability becomes the score. This runs a DeBERTa NLI model in your browser, so the text is never uploaded.

**URL:** https://convertto.tech/t/zero-shot-text-classifier
**Category:** On-Device AI (https://convertto.tech/c/local-ai-tools)
**Privacy:** Runs entirely in the browser; no upload
**Cost:** Free, no sign-up
**Last updated:** 2026-08-01

## Key facts

- **Uploads:** None — the model is downloaded to you, not the reverse
- **Runs on:** WebGPU where available, WebAssembly otherwise
- **Privacy:** Runs entirely in your browser — nothing is uploaded
- **Cost:** Free, unlimited, no sign-up

## How to use

1. Enter or paste your text.
2. Enter or paste your labels.
3. Choose the model.
4. Turn "Allow more than one label per item" on or off as needed.
5. Enter or paste your hypothesis template.
6. Press Run, then download the result when it is ready.

## FAQ

### Why does changing the hypothesis template change the answer?

Because the template is what the model actually judges. It is scoring "does this text entail: This example is a bug report?" — so a template that names your domain ("This support ticket is about {}.") gives the model context it otherwise lacks. It is the single cheapest accuracy improvement available here.

### When should I use a trained classifier instead?

Once you have a few hundred labelled examples. A fine-tuned model will beat zero-shot comfortably and run an order of magnitude faster. Zero-shot exists for the stage before you have that data — or for categories that change too often to train for.

## Sources

- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) — arXiv

## Related tools

- [AI Text Summarizer](https://convertto.tech/t/ai-text-summarizer): Condense an article into a few sentences with a summarisation model running locally — nothing uploaded, no limit.
- [Sentence Splitter](https://convertto.tech/t/sentence-splitter): Break a paragraph into one sentence per line, handling abbreviations and decimals correctly.
- [AI Alt Text Generator](https://convertto.tech/t/ai-alt-text-generator): Describe an image with a captioning model running in your browser — a first draft for the alt attribute you then edit.
- [AI Audio Transcription](https://convertto.tech/t/ai-audio-transcription): Run Whisper locally to turn speech into text, SRT, VTT or a timestamped transcript. Works offline after the first run.
- [AI Background Remover](https://convertto.tech/t/ai-background-remover): Cut the subject out of a photo with a real segmentation model that runs in your browser. No upload, no sign-up, no watermark.
- [AI Content Detector](https://convertto.tech/t/ai-content-detector): Score text with a real language model in your browser and see the per-token surprise that detection methods are built on.
- [AI Photo Redactor](https://convertto.tech/t/ai-photo-redactor): Detect people, faces, plates and screens with an object-detection model in your browser, then blur or black them out.
- [AI Text to Speech](https://convertto.tech/t/ai-text-to-speech): Turn text into natural speech with a neural voice running in your browser, and download the audio as a WAV file.