Upload 7 files

.gitattributes

@@ -53,3 +53,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.jpg filter=lfs diff=lfs merge=lfs -text
 *.jpeg filter=lfs diff=lfs merge=lfs -text
 *.webp filter=lfs diff=lfs merge=lfs -text
+bg3.ndjson filter=lfs diff=lfs merge=lfs -text
+english.loca.json filter=lfs diff=lfs merge=lfs -text

bg3.ndjson

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5737710b1d1fa0df83989d752f03dbd7b2443d44f2104ba5c2a1eab89c61326a
+size 42257474

compute_statistics.py

@@ -0,0 +1,35 @@
+import json
+
+def get_sample():
+    with open("bg3.ndjson", "r") as bg3_f:
+        for line in bg3_f:
+            yield json.loads(line.strip())
+
+def main():
+    speaker_stats = {}
+    total_duration = 0.0
+    total_num_samples = 0
+    for metadata in get_sample():
+        total_num_samples+=1
+        duration = metadata["duration"]
+        total_duration += duration
+        spk_id = metadata["spk_id"]
+        if spk_id not in speaker_stats.keys():
+            speaker_stats[spk_id] = dict(num_samples=0, total_duration=0)
+        speaker_stats[spk_id]["num_samples"] += 1
+        speaker_stats[spk_id]["total_duration"] += duration
+
+    for spk_id in speaker_stats.keys():
+        speaker_stats[spk_id]["mean_duration"] = speaker_stats[spk_id]["total_duration"]/speaker_stats[spk_id]["num_samples"]
+
+    speaker_stats["global"] = dict(
+        num_samples=total_num_samples, 
+        total_duration=total_duration, 
+        mean_duration=total_duration/len(speaker_stats)
+    )
+
+    print(json.dumps(speaker_stats, indent=2))
+
+
+if __name__=="__main__":
+    main()

create_dataset.py

@@ -0,0 +1,47 @@
+import json
+import re
+from os import getenv
+from pathlib import Path
+from tqdm import tqdm
+from soundfile import SoundFile
+
+ROOT = getenv("BG3_VOICE")
+assert ROOT is not None, "Please set the BG3_VOICE environment variable to the directory Voice.pak was unpacked in."
+
+
+def main():
+    Path("./wav/").mkdir(exist_ok=True)
+    root_path = Path(ROOT)
+    sound_path = root_path/"Mods/Gustav/Localization/English/Soundbanks"
+    with open("english.loca.json", "r") as loca_f:
+        uid_to_txt = json.load(loca_f)
+    with open("bg3.ndjson", "w") as bg3_f:
+        for file in tqdm(list(sound_path.glob(f"*.wav"))):
+            file_path=file.resolve()
+            try:
+                contentuid = re.search(r"h.*(?=\.wem)", file_path.stem).group()
+            except AttributeError:
+                continue
+            try:
+                text = uid_to_txt[contentuid]
+            except KeyError:
+                continue
+            text = re.sub(r"\(\[.*\]\)", "", text)
+            Path("wav/"+file_path.name).hardlink_to(file_path)
+            snd_file = SoundFile(str(file_path), "r")
+            sample = dict(
+                name=file_path.name,
+                text=text,
+                spk_id=file_path.name.replace(
+                    f"_{contentuid}.wem.wav",""
+                ),
+                is_effect=bool(re.match(r"v[0-9]", text)),
+                samplerate=snd_file.samplerate,
+                duration=snd_file.frames/snd_file.samplerate,
+                n_channels=snd_file.channels,
+            )
+            json.dump(sample, bg3_f)
+            bg3_f.write("\n")
+
+if __name__ == "__main__":
+    main()

english.loca.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c8a5aafd8072130ce2deab024a597017449de8245b397dcd11dac73c8fe4dd9
+size 23934356

requirements.in

@@ -0,0 +1,2 @@
+soundfile
+tqdm

requirements.txt

@@ -0,0 +1,8 @@
+# This file was autogenerated by uv via the following command:
+#    uv pip compile requirements.in -o requirements.txt
+cffi==1.16.0
+    # via soundfile
+pycparser==2.22
+    # via cffi
+soundfile==0.12.1
+tqdm==4.66.4