Orpheus TTS Dataset
This dataset is formatted for use with Orpheus TTS model training.
Dataset Information
- Format: Parquet with Audio feature
- Sample Rate: 24000Hz (Orpheus requirement)
- Columns: audio, text, filename (+ any additional metadata)
Usage
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("edfefefef23232323/richtofen_dataset")
# The audio column is automatically cast to Audio feature
print(dataset["train"][0])
Emotion Tags
This dataset may include emotion tags in angle brackets (e.g., <laugh>, <sigh>)
which are supported by Orpheus TTS for expressive speech generation.
Training with Unsloth
from unsloth import FastLanguageModel
from datasets import load_dataset
# Load model
model, tokenizer = FastLanguageModel.from_pretrained(
"unsloth/orpheus-3b-0.1-ft",
load_in_4bit=False
)
# Load your dataset
dataset = load_dataset("edfefefef23232323/richtofen_dataset", split="train")
# Continue with training...