Curated lists / Harness

Benchmark harness: Best Speech-to-Text APIs

Everything needed to reproduce the numbers: the dataset builder, the provider runner, the scorer, and the raw per-file output each provider returned. API keys are read from environment variables at run time and are not included.

Used by: Best Speech-to-Text APIs

• build_dataset.py3,404 B
• results/2026-09-10/assemblyai/ami-ES2002a-0-600.json6,004 B
• results/2026-09-10/assemblyai/libri-1089-134686.json4,244 B
• results/2026-09-10/assemblyai/libri-4507-16021.json5,710 B
• results/2026-09-10/deepgram/ami-ES2002a-0-600.json6,048 B
• results/2026-09-10/deepgram/libri-1089-134686.json4,200 B
• results/2026-09-10/deepgram/libri-4507-16021.json5,600 B
• results/2026-09-10/gladia/ami-ES2002a-0-600.json5,737 B
• results/2026-09-10/gladia/libri-1089-134686.json4,161 B
• results/2026-09-10/gladia/libri-4507-16021.json5,457 B
• results/2026-09-10/results.json5,455 B
• results/2026-09-10/speechmatics/ami-ES2002a-0-600.json6,502 B
• results/2026-09-10/speechmatics/libri-1089-134686.json4,221 B
• results/2026-09-10/speechmatics/libri-4507-16021.json5,599 B
• results/2026-09-10/transcribe/ami-ES2002a-0-600.json5,936 B
• results/2026-09-10/transcribe/libri-1089-134686.json4,203 B
• results/2026-09-10/transcribe/libri-4507-16021.json5,577 B
• results/2026-09-10/whisper/ami-ES2002a-0-600.json5,437 B
• results/2026-09-10/whisper/libri-1089-134686.json4,122 B
• results/2026-09-10/whisper/libri-4507-16021.json5,898 B
• run.py8,652 B
• score.py3,262 B