Emotion tracking and analysis platform supporting 200+ languages. We convert, quantize, and publish production-ready ONNX models for in-browser inference with transformers.js — zero server cost, zero latency, complete privacy.
| Model | Task | Size |
|---|---|---|
| roberta-base-go-emotions-onnx | Emotion detection (28 labels) | ~124 MB |
| distilbert-base-uncased-emotion-onnx | Emotion classification (6 labels) | ~67 MB |
| emotion-english-distilroberta-base-onnx | Emotion classification (7 labels) | ~82 MB |
| twitter-roberta-base-sentiment-onnx | Sentiment analysis (3-class) | ~124 MB |
| Model | Task | Size |
|---|---|---|
| deberta-v3-base-mnli-onnx | Zero-shot classification | ~243 MB |
| bart-large-cnn-onnx | Text summarization | ~493 MB |
| Model | Task | Size |
|---|---|---|
| nllb-200-1.3B-onnx | Translation (200 languages, full) | ~1.81 GB |
| nllb-200-distilled-1.3B-onnx | Translation (200 languages, distilled) | ~1.81 GB |
All packages available on npm under @affectively: