🎙️ OuteTTS Saudi Arabic - LoRA Adapter (Step 1000)
Fine-tuned LoRA adapter for Saudi Arabic text-to-speech.
Model Details
- Base Model: OuteAI/OuteTTS-0.2-500M
- Language: Arabic (Saudi Dialect)
- Training Steps: 1000
- Training Samples: 7,400
- Method: LoRA (Low-Rank Adaptation)
Usage
from transformers import AutoModelForCausalLM
from peft import PeftModel
# Load base model
base_model = AutoModelForCausalLM.from_pretrained("OuteAI/OuteTTS-0.2-500M")
# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "ISTNetworks/outerTTS-saudi-lora-1000")
tokenizer = AutoTokenizer.from_pretrained("ISTNetworks/outerTTS-saudi-lora-1000")
# Use for inference
Checkpoint Series
This is checkpoint 1000 in a series of training checkpoints:
- checkpoint-1000 (early stage)
- checkpoint-2000 (mid-training)
- checkpoint-3000 (advanced)
- checkpoint-4000 (mature)
Use Cases
- Banking IVR systems
- Saudi Arabic voice assistants
- Customer service automation
License
Apache 2.0
Model tree for ISTNetworks/outerTTS-saudi-lora-1000
Base model
OuteAI/OuteTTS-0.2-500M
import outetts enum = outetts.Models("ISTNetworks/outerTTS-saudi-lora-1000".split("/", 1)[1]) # VERSION_1_0_SIZE_1B cfg = outetts.ModelConfig.auto_config(enum, outetts.Backend.HF) tts = outetts.Interface(cfg) speaker = tts.load_default_speaker("EN-FEMALE-1-NEUTRAL") tts.generate( outetts.GenerationConfig( text="Hello there, how are you doing?", speaker=speaker, ) ).save("output.wav")