Create README.md

README.md

@@ -0,0 +1,27 @@
+---
+task_categories:
+- text-to-speech
+language:
+- en
+pretty_name: CommonVoice Clones
+size_categories:
+- 1K<n<10K
+---
+
+# CommonVoice Clones
+
+This dataset consists of recordings taken from the CommonVoice english dataset. 
+Each voice and transcript are used as input to a voice cloner, and generate a cloned version of the voice and text.
+
+## TTS Models
+We use the following high-scoring models from the TTS leaderboard:
+
+- playHT
+- metavoice
+- StyleTTSv2
+- XttsV2
+
+## Data Sampling
+
+The subset of the data was chosen such that a balanced amount of male/female voices are included.
+We also include a suggested train, validation and test split. Unique voice IDs do not overlap between the splits, i.e. a voice used to clone a sample in train will never be found in validation and test.