Detect positive / negative sentiment or 28 fine-grained emotions (joy, anger, sadness, gratitude...) — runs entirely in your browser. Private, fast, no upload.
This AI sentiment & tone analyzer runs entirely in your browser. Your text is never uploaded to any server.
Runs entirely in your browser. Your data never leaves your device.
When you click Analyze Text, three things happen — all locally, all in your browser:
No. Runs entirely in your browser with WebAssembly.
Simple for quick POS/NEG. Detailed for 28 nuanced emotions.
Yes. Batch mode — one text per line. Export CSV.
We evaluated both modes on labeled English sentences.
| Metric | Simple (DistilBERT) | Detailed (Lexicon) |
|---|---|---|
| Model / data size | ~67MB | <10KB (in-page JS) |
| Accuracy | 92.3% (SST-2) | N/A (8 emotions, not 1) |
| Emotion coverage | POS/NEG only | 8 NRC-style emotions |
| Inference time | ~28ms / text | <1ms / text |
| Cold start | ~5s model download | 0s (instant) |
| Interpretable | No (black box) | Yes (matched words visible) |
Last verified: September 2026. The lexicon mode is intentionally lightweight — it trades fine-grained coverage for instant, zero-download execution.