Anubodh/polyglot-voice-bench

ci: update HF Space README

voice benchmark abstract

fix: replace deprecated width="stretch" with use_container_width in dataframes

updated readme

ci: retrigger HF deploy

ci: opt into Node.js 24 for GitHub Actions runners

fix: update HF Space URL to correct username (Anubodh)

UI: Project Story

feat(v4): add Streamlit leaderboard app for HF Space

feat(scripts): add split_recording.py to split long recording by silence

feat(v2b+v3): AssemblyAI Universal-2 results + own-voice dataset infra

feat(v2b+v3): add AssemblyAI Universal-2 provider and own-voice dataset infra

feat(v3): add own-voice dataset infrastructure + recording protocol

feat(v2): add AssemblyAI Universal-3 Pro + English baseline, 3-way comparison

feat(v2): add AssemblyAI Universal-3 Pro + English baseline for 3-way comparison

feat(v1): add Whisper Large v3 via Groq for cross-provider comparison

docs: document v0 findings and roadmap to v1

feat(scripts): add run_v0 end-to-end entry point

feat(runner): add resumable benchmark runner with per-sample persistence

feat(providers): add Deepgram Nova-3 Multilingual provider

feat(datasets): add FLEURS loader for Hindi and Nepali subsets

feat(metrics): add WER, CER, error breakdown with bootstrap CIs + tests

feat(types): add TranscriptionResult and AudioSample dataclasses

chore: scaffold project structure with uv, Makefile, and stub docs