app.py

import gradio as gr
import spaces

import os
import shutil
os.environ['TOKENIZERS_PARALLELISM'] = 'true'
os.environ['SPCONV_ALGO'] = 'native'
from typing import *
import torch
import numpy as np
import imageio
from easydict import EasyDict as edict
from trellis.pipelines import TrellisTextTo3DPipeline
from trellis.representations import Gaussian, MeshExtractResult
from trellis.utils import render_utils, postprocessing_utils
import joblib # Added for saving/loading state

import traceback
import sys


MAX_SEED = np.iinfo(np.int32).max
TMP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'tmp')
os.makedirs(TMP_DIR, exist_ok=True)


def start_session(req: gr.Request):
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)
    
    
def end_session(req: gr.Request):
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    shutil.rmtree(user_dir)


def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
    return {
        'gaussian': {
            **gs.init_params,
            '_xyz': gs._xyz.cpu().numpy(),
            '_features_dc': gs._features_dc.cpu().numpy(),
            '_scaling': gs._scaling.cpu().numpy(),
            '_rotation': gs._rotation.cpu().numpy(),
            '_opacity': gs._opacity.cpu().numpy(),
        },
        'mesh': {
            'vertices': mesh.vertices.cpu().numpy(),
            'faces': mesh.faces.cpu().numpy(),
        },
    }
    
    
def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
    gs = Gaussian(
        aabb=state['gaussian']['aabb'],
        sh_degree=state['gaussian']['sh_degree'],
        mininum_kernel_size=state['gaussian']['mininum_kernel_size'],
        scaling_bias=state['gaussian']['scaling_bias'],
        opacity_bias=state['gaussian']['opacity_bias'],
        scaling_activation=state['gaussian']['scaling_activation'],
    )
    gs._xyz = torch.tensor(state['gaussian']['_xyz'], device='cuda')
    gs._features_dc = torch.tensor(state['gaussian']['_features_dc'], device='cuda')
    gs._scaling = torch.tensor(state['gaussian']['_scaling'], device='cuda')
    gs._rotation = torch.tensor(state['gaussian']['_rotation'], device='cuda')
    gs._opacity = torch.tensor(state['gaussian']['_opacity'], device='cuda')
    
    mesh = edict(
        vertices=torch.tensor(state['mesh']['vertices'], device='cuda'),
        faces=torch.tensor(state['mesh']['faces'], device='cuda'),
    )
    
    return gs, mesh


def get_seed(randomize_seed: bool, seed: int) -> int:
    """
    Get the random seed.
    """
    return np.random.randint(0, MAX_SEED) if randomize_seed else seed


@spaces.GPU
def text_to_3d(
    prompt: str,
    seed: int,
    ss_guidance_strength: float,
    ss_sampling_steps: int,
    slat_guidance_strength: float,
    slat_sampling_steps: int,
    req: gr.Request,
) -> Tuple[str, str, str]:
    """
    Convert an text prompt to a 3D model.
    Args:
        prompt (str): The text prompt.
        seed (int): The random seed.
        ss_guidance_strength (float): The guidance strength for sparse structure generation.
        ss_sampling_steps (int): The number of sampling steps for sparse structure generation.
        slat_guidance_strength (float): The guidance strength for structured latent generation.
        slat_sampling_steps (int): The number of sampling steps for structured latent generation.
    Returns:
        str: Path to the saved state file.
        str: Path to the generated video.
        str: Path to the saved state file (for internal buffer).
    """
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)
    outputs = pipeline.run(
        prompt,
        seed=seed,
        formats=["gaussian", "mesh"],
        sparse_structure_sampler_params={
            "steps": ss_sampling_steps,
            "cfg_strength": ss_guidance_strength,
        },
        slat_sampler_params={
            "steps": slat_sampling_steps,
            "cfg_strength": slat_guidance_strength,
        },
    )
    video = render_utils.render_video(outputs['gaussian'][0], num_frames=120)['color']
    video_geo = render_utils.render_video(outputs['mesh'][0], num_frames=120)['normal']
    video = [np.concatenate([video[i], video_geo[i]], axis=1) for i in range(len(video))]
    video_path = os.path.join(user_dir, 'sample.mp4')
    imageio.mimsave(video_path, video, fps=15)
    state = pack_state(outputs['gaussian'][0], outputs['mesh'][0])
    
    # Save state to file
    state_file_path = os.path.join(user_dir, f'state_{seed}.joblib')
    try:
        joblib.dump(state, state_file_path)
        print(f"[Trellis] State saved to {state_file_path}")
    except Exception as e:
        print(f"Error saving state to {state_file_path}: {e}")
        # Decide how to handle error - maybe return None or raise?
        # For now, let's allow it to proceed but log the error
        state_file_path = None # Indicate failure

    torch.cuda.empty_cache()
    # Return state file path for API, video path for Video, and state path again for internal buffer
    # Return None for path if saving failed
    return state_file_path, video_path, state_file_path


@spaces.GPU(duration=90)
def extract_glb(
    state_file_path: str, # Changed input from state: dict
    mesh_simplify: float,
    texture_size: int,
    req: gr.Request,
) -> Tuple[str, str]:
    """
    Extract a GLB file from the 3D model state file.
    Args:
        state_file_path (str): Path to the file containing the state.
        mesh_simplify (float): The mesh simplification factor.
        texture_size (int): The texture resolution.
    Returns:
        str: The path to the extracted GLB file.
        str: The path to the extracted GLB file (for download button).
    """
    if not state_file_path or not os.path.exists(state_file_path):
        print(f"Error: State file path invalid or file not found: {state_file_path}")
        # Return dummy paths or raise an error
        return None, None 

    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)
    
    # Load state from file
    try:
        state = joblib.load(state_file_path)
        print(f"[Trellis] State loaded from {state_file_path}")
    except Exception as e:
        print(f"Error loading state from {state_file_path}: {e}")
        return None, None

    gs, mesh = unpack_state(state)
    glb = postprocessing_utils.to_glb(gs, mesh, simplify=mesh_simplify, texture_size=texture_size, verbose=False)
    glb_path = os.path.join(user_dir, 'sample.glb')
    glb.export(glb_path)
    torch.cuda.empty_cache()

    # Optional: Clean up the state file after use
    try:
        os.remove(state_file_path)
        print(f"[Trellis] Cleaned up state file: {state_file_path}")
    except OSError as e:
        print(f"Error removing state file {state_file_path}: {e.strerror}")

    return glb_path, glb_path


@spaces.GPU
def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
    """
    Extract a Gaussian file from the 3D model.
    Args:
        state (dict): The state of the generated 3D model.
    Returns:
        str: The path to the extracted Gaussian file.
    """
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)
    gs, _ = unpack_state(state)
    gaussian_path = os.path.join(user_dir, 'sample.ply')
    gs.save_ply(gaussian_path)
    torch.cuda.empty_cache()
    return gaussian_path, gaussian_path


output_buf = gr.State()
video_output = gr.Video(label="Generated 3D Asset", autoplay=True, loop=True, height=300)
model_output = gr.Model3D(label="Extracted GLB/Gaussian", height=300)

# Change hidden JSON to hidden Textbox for the state file path
state_output_path_textbox = gr.Textbox(visible=False, label="State File Path Output")

with gr.Blocks(delete_cache=(600, 600)) as demo:
    gr.Markdown("""
    ## Text to 3D Asset with [TRELLIS](https://trellis3d.github.io/)
    * Type a text prompt and click "Generate" to create a 3D asset.
    * If you find the generated 3D asset satisfactory, click "Extract GLB" to extract the GLB file and download it.
    """)
    
    with gr.Row():
        with gr.Column():
            text_prompt = gr.Textbox(label="Text Prompt", lines=5)
            
            with gr.Accordion(label="Generation Settings", open=False):
                seed = gr.Slider(0, MAX_SEED, label="Seed", value=0, step=1)
                randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
                gr.Markdown("Stage 1: Sparse Structure Generation")
                with gr.Row():
                    ss_guidance_strength = gr.Slider(0.0, 10.0, label="Guidance Strength", value=7.5, step=0.1)
                    ss_sampling_steps = gr.Slider(1, 50, label="Sampling Steps", value=25, step=1)
                gr.Markdown("Stage 2: Structured Latent Generation")
                with gr.Row():
                    slat_guidance_strength = gr.Slider(0.0, 10.0, label="Guidance Strength", value=7.5, step=0.1)
                    slat_sampling_steps = gr.Slider(1, 50, label="Sampling Steps", value=25, step=1)

            generate_btn = gr.Button("Generate")
            
            with gr.Accordion(label="GLB Extraction Settings", open=False):
                mesh_simplify = gr.Slider(0.9, 0.98, label="Simplify", value=0.95, step=0.01)
                texture_size = gr.Slider(512, 2048, label="Texture Size", value=1024, step=512)
            
            with gr.Row():
                extract_glb_btn = gr.Button("Extract GLB", interactive=False)
                extract_gs_btn = gr.Button("Extract Gaussian", interactive=False)
            gr.Markdown("""
                        *NOTE: Gaussian file can be very large (~50MB), it will take a while to display and download.*
                        """)

        with gr.Column():
            video_output = gr.Video(label="Generated 3D Asset", autoplay=True, loop=True, height=300)
            model_output = gr.Model3D(label="Extracted GLB/Gaussian", height=300)
            
            with gr.Row():
                download_glb = gr.DownloadButton(label="Download GLB", interactive=False)
                download_gs = gr.DownloadButton(label="Download Gaussian", interactive=False)  
    
    output_buf = gr.State()

    # Handlers
    demo.load(start_session)
    demo.unload(end_session)

    generate_btn.click(
        get_seed,
        inputs=[randomize_seed, seed],
        outputs=[seed],
    ).then(
        text_to_3d,
        inputs=[text_prompt, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps],
        # Output state path to hidden Textbox, video to Video, state path to internal buffer
        outputs=[state_output_path_textbox, video_output, output_buf],
    ).then(
        lambda: tuple([gr.Button(interactive=True), gr.Button(interactive=True)]),
        outputs=[extract_glb_btn, extract_gs_btn],
    )

    video_output.clear(
        lambda: tuple([gr.Button(interactive=False), gr.Button(interactive=False)]),
        outputs=[extract_glb_btn, extract_gs_btn],
    )

    extract_glb_btn.click(
        extract_glb,
        # Input state path from internal buffer (assuming it holds the path now)
        inputs=[output_buf, mesh_simplify, texture_size],
        outputs=[model_output, download_glb],
    ).then(
        lambda: gr.Button(interactive=True),
        outputs=[download_glb],
    )
    
    extract_gs_btn.click(
        extract_gaussian,
        # This likely needs adjustment too if it relies on output_buf holding the state dict
        inputs=[output_buf], 
        outputs=[model_output, download_gs],
    ).then(
        lambda: gr.Button(interactive=True),
        outputs=[download_gs],
    )

    model_output.clear(
        lambda: gr.Button(interactive=False),
        outputs=[download_glb],
    )
    

# Launch the Gradio app
if __name__ == "__main__":
    pipeline = TrellisTextTo3DPipeline.from_pretrained("JeffreyXiang/TRELLIS-text-xlarge")
    pipeline.cuda()
    demo.launch()

# --- API-only endpoint for server integration ---
# This exposes text_to_3d with gr.JSON() as the first output, so the state object is included in the API response.
# Not wired to the UI; use for API calls only.
api_text_to_3d = gr.Interface(
    fn=lambda prompt, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps: text_to_3d(
        prompt, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps, gr.Request()
    ),
    inputs=[
        gr.Textbox(label="Text Prompt"),
        gr.Number(label="Seed"),
        gr.Number(label="SS Guidance Strength"),
        gr.Number(label="SS Sampling Steps"),
        gr.Number(label="SLAT Guidance Strength"),
        gr.Number(label="SLAT Sampling Steps"),
    ],
    # Note: This API is technically available but not used by the server.
    # The server uses the main UI endpoint modified to return JSON first.
    outputs=[gr.JSON(label="State Object"), gr.Textbox(label="Video Path")],
    allow_flagging="never",
    description="API endpoint for text_to_3d that returns the state object as JSON. Not for UI use.",
)

# --- API-only endpoint for GLB extraction --- 
# Explicitly defines state input as JSON for server calls.
api_extract_glb = gr.Interface(
    fn=lambda state_file_path, mesh_simplify, texture_size: extract_glb(
        state_file_path, mesh_simplify, texture_size, gr.Request()
    ),
    inputs=[
        gr.Textbox(label="State File Path"), # Expect state file path as string
        gr.Slider(0.9, 0.98, label="Simplify", value=0.95, step=0.01),
        gr.Slider(512, 2048, label="Texture Size", value=1024, step=512)
    ],
    # Corresponds to the tuple returned by extract_glb: (glb_path, glb_path)
    outputs=[
        gr.Model3D(label="Extracted GLB Path Output"), # Maps to first glb_path
        gr.File(label="Downloadable GLB File") # Maps to second glb_path 
    ],
    allow_flagging="never",
    description="API endpoint for extract_glb that accepts state as JSON. Not for UI use.",
)