app.py

import torch
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
from huggingface_hub import snapshot_download
from snac import SNAC
import time  # Import the time module
from dotenv import load_dotenv
from optimum.bettertransformer import BetterTransformer

load_dotenv()


# Check if CUDA is available, otherwise use CPU
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

# 1. Load SNAC Model (for audio decoding)
print("Loading SNAC model...")
snac_model = SNAC.from_pretrained("hubertsiuzdak/snac_24khz")
snac_model = snac_model.to(device)
snac_model.eval()  # Set SNAC to evaluation mode

# 2. Load Orpheus Language Model (for text-to-token generation)
model_name = "canopylabs/orpheus-3b-0.1-ft"

# Download only necessary files (config and safetensors)
print("Downloading Orpheus model files...")
snapshot_download(
    repo_id=model_name,
    allow_patterns=[
        "config.json",
        ".safetensors",
        "model.safetensors.index.json",
    ],
    ignore_patterns=[
        "optimizer.pt",
        "pytorch_model.bin",
        "training_args.bin",
        "scheduler.pt",
        "tokenizer.json",
        "tokenizer_config.json",
        "special_tokens_map.json",
        "vocab.json",
        "merges.txt",
        "tokenizer."
    ]
)

print("Loading Orpheus model...")
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True)

# --- Optimization 1: Convert to BetterTransformer ---
try:
    model = BetterTransformer.transform(model)
    print("Model converted to BetterTransformer for faster inference.")
except Exception as e:
    print(f"BetterTransformer conversion failed: {e}. Proceeding without it.")

model.to(device)
model.eval()  # Set the Orpheus model to evaluation mode
tokenizer = AutoTokenizer.from_pretrained(model_name)
print(f"Orpheus model loaded to {device}")


# --- Function Definitions ---

def process_prompt(prompt, voice, tokenizer, device):
    """Processes the text prompt and converts it to input IDs."""
    prompt = f"{voice}: {prompt}"
    input_ids = tokenizer(prompt, return_tensors="pt").input_ids

    start_token = torch.tensor([[128259]], dtype=torch.int64)  # Start of human
    end_tokens = torch.tensor([[128009, 128260]], dtype=torch.int64)  # End of text, End of human

    modified_input_ids = torch.cat([start_token, input_ids, end_tokens], dim=1)  # SOH SOT Text EOT EOH

    # No padding needed for single input
    attention_mask = torch.ones_like(modified_input_ids)

    return modified_input_ids.to(device), attention_mask.to(device)

def parse_output(generated_ids):
    """Parses the generated token IDs to extract the audio codes."""
    token_to_find = 128257  # SOT token
    token_to_remove = 128258 # EOT token

    token_indices = (generated_ids == token_to_find).nonzero(as_tuple=True)

    if len(token_indices[1]) > 0:
        last_occurrence_idx = token_indices[1][-1].item()
        cropped_tensor = generated_ids[:, last_occurrence_idx + 1:]
    else:
        cropped_tensor = generated_ids

    processed_rows = []
    for row in cropped_tensor:
        masked_row = row[row != token_to_remove]
        processed_rows.append(masked_row)

    code_lists = []
    for row in processed_rows:
        row_length = row.size(0)
        new_length = (row_length // 7) * 7  # Ensure divisibility by 7
        trimmed_row = row[:new_length]
        trimmed_row = [t - 128266 for t in trimmed_row]  # Adjust code values
        code_lists.append(trimmed_row)

    return code_lists[0]  # Return codes for the first (and only) sequence


def redistribute_codes(code_list, snac_model):
    """Redistributes the audio codes into the format required by SNAC."""
    device = next(snac_model.parameters()).device  # Get the device of SNAC model

    layer_1 = []
    layer_2 = []
    layer_3 = []
    for i in range(len(code_list) // 7):  # Corrected loop condition
        layer_1.append(code_list[7*i])
        layer_2.append(code_list[7*i+1]-4096)
        layer_3.append(code_list[7*i+2]-(2*4096))
        layer_3.append(code_list[7*i+3]-(3*4096))
        layer_2.append(code_list[7*i+4]-(4*4096))
        layer_3.append(code_list[7*i+5]-(5*4096))
        layer_3.append(code_list[7*i+6]-(6*4096))

    # Move tensors to the same device as the SNAC model
    codes = [
        torch.tensor(layer_1, device=device).unsqueeze(0),
        torch.tensor(layer_2, device=device).unsqueeze(0),
        torch.tensor(layer_3, device=device).unsqueeze(0)
    ]

    audio_hat = snac_model.decode(codes)
    return audio_hat.detach().squeeze().cpu().numpy()  # Return CPU numpy array


def generate_speech(text, voice, temperature, top_p, repetition_penalty, max_new_tokens, progress=gr.Progress()):
    """Generates speech from the given text using Orpheus and SNAC."""
    if not text.strip():
        return None

    try:
        start_time = time.time()  # Start timing

        progress(0.1, "Processing text...")
        input_ids, attention_mask = process_prompt(text, voice, tokenizer, device)
        process_time = time.time() - start_time
        print(f"Text processing time: {process_time:.2f} seconds")

        start_time = time.time()  # Reset timer
        progress(0.3, "Generating speech tokens...")
        with torch.no_grad():
            generated_ids = model.generate(
                input_ids=input_ids,
                attention_mask=attention_mask,
                max_new_tokens=max_new_tokens,
                do_sample=True,
                temperature=temperature,
                top_p=top_p,
                repetition_penalty=repetition_penalty,
                num_return_sequences=1,
                eos_token_id=128258,
            )
        generation_time = time.time() - start_time
        print(f"Token generation time: {generation_time:.2f} seconds")

        start_time = time.time()  # Reset timer
        progress(0.6, "Processing speech tokens...")
        code_list = parse_output(generated_ids)
        parse_time = time.time() - start_time
        print(f"Token parsing time: {parse_time:.2f} seconds")


        start_time = time.time()  # Reset timer
        progress(0.8, "Converting to audio...")
        audio_samples = redistribute_codes(code_list, snac_model)
        audio_time = time.time() - start_time
        print(f"Audio conversion time: {audio_time:.2f} seconds")

        return (24000, audio_samples)  # Return sample rate and audio
    except Exception as e:
        print(f"Error generating speech: {e}")
        return None



# --- Gradio Interface Setup ---

examples = [
    ["Hey there my name is Tara, <chuckle> and I'm a speech generation model that can sound like a person.", "tara", 0.6, 0.95, 1.1, 1200],
    ["I've also been taught to understand and produce paralinguistic things like sighing, or chuckling, or yawning!", "dan", 0.7, 0.95, 1.1, 1200],
    ["I live in San Francisco, and have, uhm let's see, 3 billion 7 hundred ... well, lets just say a lot of parameters.", "emma", 0.6, 0.9, 1.2, 1200]
]

VOICES = ["tara", "dan", "josh", "emma"]

with gr.Blocks(title="Orpheus Text-to-Speech") as demo:
    gr.Markdown("""
    # 🎵 Orpheus Text-to-Speech
    Enter text below to convert to speech.
    """)
    with gr.Row():
        with gr.Column(scale=3):
            text_input = gr.Textbox(
                label="Text to speak",
                placeholder="Enter your text here...",
                lines=5
            )
            voice = gr.Dropdown(
                choices=VOICES,
                value="tara",
                label="Voice"
            )

            with gr.Accordion("Advanced Settings", open=False):
                temperature = gr.Slider(
                    minimum=0.1, maximum=1.5, value=0.6, step=0.05,
                    label="Temperature"
                )
                top_p = gr.Slider(
                    minimum=0.1, maximum=1.0, value=0.95, step=0.05,
                    label="Top P"
                )
                repetition_penalty = gr.Slider(
                    minimum=1.0, maximum=2.0, value=1.1, step=0.05,
                    label="Repetition Penalty"
                )
                max_new_tokens = gr.Slider(
                    minimum=100, maximum=2000, value=1200, step=100,
                    label="Max Length"
                )

            with gr.Row():
                submit_btn = gr.Button("Generate Speech", variant="primary")
                clear_btn = gr.Button("Clear")

        with gr.Column(scale=2):
            audio_output = gr.Audio(label="Generated Speech", type="numpy")

    gr.Examples(
        examples=examples,
        inputs=[text_input, voice, temperature, top_p, repetition_penalty, max_new_tokens],
        outputs=audio_output,
        fn=generate_speech,
        cache_examples=True,
    )

    submit_btn.click(
        fn=generate_speech,
        inputs=[text_input, voice, temperature, top_p, repetition_penalty, max_new_tokens],
        outputs=audio_output
    )

    clear_btn.click(
        fn=lambda: (None, None),
        inputs=[],
        outputs=[text_input, audio_output]
    )

if __name__ == "__main__":
    demo.queue().launch(share=False)