Create README.md

README.md

@@ -0,0 +1,55 @@
+---
+task_categories:
+- text-to-speech
+language:
+- en
+pretty_name: bg3
+size_categories:
+- 100K<n<1M
+---
+
+Voicelines extracted from BG3
+
+You'll find the code for reproducing it in the files.
+
+
+Untar the audio files (`tar xvf bg3.tar.gz`)
+
+The metadata looks like this usually;
+```json
+{
+  "name": "v4316b95f66244601bcebbdddb8e0ba62_h0ad754d9gc97eg4fe9g9ae3gc0c2a2b3668c.wem.wav",
+  "text": "A table somewhere. For one.",
+  "spk_id": "v4316b95f66244601bcebbdddb8e0ba62",
+  "is_effect": false,
+  "samplerate": 48000,
+  "duration": 2.7868958333333333,
+  "n_channels": 1
+}
+```
+but might also have those.
+```json
+{
+  "name": "vaa0a8ffc4ab34aae82166d61f1619fdc_h843a4ee4g144ag48d8g8f99g32f357d736d5.wem.wav",
+  "text": "<i>Deaddeaddead.</i>",
+  "spk_id": "vaa0a8ffc4ab34aae82166d61f1619fdc",
+  "is_effect": false,
+  "samplerate": 48000,
+  "duration": 1.313375,
+  "n_channels": 1
+}
+```
+There are also some voice effects.
+```json
+{
+  "name": "v3ed74f063c6042dc83f6f034cb47c679_he1904e3fg0dfag47c1gbaa2g43752a0a229b.wem.wav",
+  "text": "Closed Mouth Effort v12",
+  "spk_id": "v3ed74f063c6042dc83f6f034cb47c679",
+  "is_effect": true,
+  "samplerate": 48000,
+  "duration": 0.3889375,
+  "n_channels": 1
+}
+```
+
+Warning: some files have 2 channels, do not assume they're all mono.