Subs (subs.alftech.space) — what runs in your browser and whose it is
=====================================================================

Everything below is served from this site. Nothing is fetched from
huggingface.co, a CDN or anyone else while the page runs.

Speech model: Whisper, by OpenAI
  Paper:    Radford et al., "Robust Speech Recognition via Large-Scale Weak
            Supervision", 2022.
  Code and weights released under the MIT licence, see LICENSE-whisper.txt
  (https://github.com/openai/whisper). The Hugging Face model cards for
  openai/whisper-base and openai/whisper-small list Apache-2.0; the original
  release is MIT, and both permit this use.

  ONNX conversion and int8 quantisation: onnx-community (Hugging Face),
  Apache-2.0.

  whisper-base/   "Fast"      onnx-community/whisper-base
                              revision 27a07b4b52b3a4c5886134bb4320e93e1a4bbc7c
      onnx/encoder_model_quantized.onnx          23,201,314 bytes
      onnx/decoder_model_merged_quantized.onnx   53,693,315 bytes
      + config, generation config, preprocessor and tokenizer JSON (2.8 MB)

  whisper-small/  "Accurate"  onnx-community/whisper-small
                              revision 36050c46d777d46dc4b5f43f6d90574fc38f8732
      onnx/encoder_model_quantized.onnx          92,326,160 bytes
      onnx/decoder_model_merged_quantized.onnx  156,750,845 bytes
      + the same JSON files (2.8 MB)

  sha256 of every ONNX file: tools/fetch-models.sh in the source.

Runtime
  transformers.js 4.3.1 (Hugging Face), Apache-2.0 — vendor/transformers.min.js,
    licence in vendor/LICENSE-transformers.js.txt
  ONNX Runtime Web 1.31.0-dev.20260914 (Microsoft), MIT — vendor/ort-wasm-*,
    licence in vendor/LICENSE-onnxruntime.txt

Fonts
  Manrope (SIL Open Font Licence 1.1), DM Mono (SIL Open Font Licence 1.1).