Pepper browser companion

Paper silhouette, fibers and eyebrow-free expressive eyes adapted from the owner's Pepper Android 1.15 PaperPepperRenderer. All 36 Echo electronic gestures are unchanged from the owner's bank 3.0-playful-squeaks.

Quick tone model: audeering/wav2small, revision 6ea3fe5be073ba13083bb60d14af2736dc258de9, 15,030 parameters. CC-BY-NC-SA-4.0 (noncommercial/share-alike). Model, original provenance and export parity receipt are in /tone/. License: /licenses/wav2small-CC-BY-NC-SA-4.0.txt. Model estimates arousal/dominance/valence; these are not categorical emotion probabilities.

ONNX Runtime Web: Microsoft, MIT. Transformers.js: Hugging Face, Apache-2.0. Runtime dependency notices retained in build outputs and package lock.

Optional local speech model: onnx-community/whisper-tiny.en (OpenAI Whisper tiny English; MIT), downloaded only when requested from Hugging Face. Optional semantic emotion model: onnx-community/emotion-english-distilroberta-base-ONNX. Attribution: Jochen Hartmann, Emotion English DistilRoBERTa-base, 2022. Its model card does not declare a license in repository metadata; follow the upstream model card and terms. Model weights are fetched directly by the browser, not redistributed on this static host. Model cards:
https://huggingface.co/onnx-community/whisper-tiny.en
https://huggingface.co/onnx-community/emotion-english-distilroberta-base-ONNX
https://huggingface.co/audeering/wav2small

This static host receives ordinary asset requests only. Inference is local. No audio upload, browser-vendor speech API, remote Pepper service, account or paid API is used. Preferences alone are stored locally. Model downloads are cached by the inference library/browser; use browser site-data controls to clear them.