app.py

# Version: Add API State Fix (2025-05-04)
# Changes:
# - Modified `text_to_3d` to explicitly return the serializable `state_dict` from `pack_state`
#   as the first return value. This ensures the dictionary is available via the API.
# - Modified `extract_glb` to accept `state_dict: dict` as its first argument instead of
#   relying on the implicit `gr.State` object type when called via API.
# - Kept Gradio UI bindings (`outputs=[output_buf, ...]`, `inputs=[output_buf, ...]`)
#   so the UI continues to function by passing the dictionary through output_buf.

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 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))
    # Add safety check before removing
    if os.path.exists(user_dir):
      try:
          shutil.rmtree(user_dir)
      except OSError as e:
          print(f"Error removing tmp directory {user_dir}: {e.strerror}", file=sys.stderr)


def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
    # Ensure tensors are on CPU and converted to numpy before returning the dict
    return {
        'gaussian': {
            **gs.init_params,
            '_xyz': gs._xyz.detach().cpu().numpy(),
            '_features_dc': gs._features_dc.detach().cpu().numpy(),
            '_scaling': gs._scaling.detach().cpu().numpy(),
            '_rotation': gs._rotation.detach().cpu().numpy(),
            '_opacity': gs._opacity.detach().cpu().numpy(),
        },
        'mesh': {
            'vertices': mesh.vertices.detach().cpu().numpy(),
            'faces': mesh.faces.detach().cpu().numpy(),
        },
    }


def unpack_state(state_dict: dict) -> Tuple[Gaussian, edict]:
    # Ensure the device is correctly set when unpacking
    device = 'cuda' if torch.cuda.is_available() else 'cpu'

    gs = Gaussian(
        aabb=state_dict['gaussian']['aabb'],
        sh_degree=state_dict['gaussian']['sh_degree'],
        mininum_kernel_size=state_dict['gaussian']['mininum_kernel_size'],
        scaling_bias=state_dict['gaussian']['scaling_bias'],
        opacity_bias=state_dict['gaussian']['opacity_bias'],
        scaling_activation=state_dict['gaussian']['scaling_activation'],
    )
    gs._xyz = torch.tensor(state_dict['gaussian']['_xyz'], device=device)
    gs._features_dc = torch.tensor(state_dict['gaussian']['_features_dc'], device=device)
    gs._scaling = torch.tensor(state_dict['gaussian']['_scaling'], device=device)
    gs._rotation = torch.tensor(state_dict['gaussian']['_rotation'], device=device)
    gs._opacity = torch.tensor(state_dict['gaussian']['_opacity'], device=device)

    mesh = edict(
        vertices=torch.tensor(state_dict['mesh']['vertices'], device=device),
        faces=torch.tensor(state_dict['mesh']['faces'], device=device),
    )

    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[dict, str]: # <- Changed return annotation for clarity
    """
    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:
        dict: The *serializable dictionary* representing the state of the generated 3D model. <-- CHANGE
        str: The path to the video preview of the 3D model.
    """
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)

    # --- Generation Pipeline ---
    outputs = pipeline.run(
        prompt,
        seed=seed,
        formats=["gaussian", "mesh"], # Ensure both are generated
        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,
        },
    )

    # --- Create Serializable State Dictionary --- VITAL CHANGE for API
    # Instead of returning the raw state object, return a serializable dictionary
    # which can be passed via the API correctly.
    state_dict = pack_state(outputs['gaussian'][0], outputs['mesh'][0])

    # --- Render Video Preview ---
    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)

    torch.cuda.empty_cache()

    # --- Return Serializable Dictionary and Video Path --- VITAL CHANGE for API
    return state_dict, video_path


@spaces.GPU(duration=90)
def extract_glb(
    state_dict: dict, # <-- VITAL CHANGE: Accept the dictionary directly
    mesh_simplify: float,
    texture_size: int,
    req: gr.Request,
) -> Tuple[str, str]:
    """
    Extract a GLB file from the 3D model state dictionary.
    Args:
        state_dict (dict): The serializable dictionary state of the generated 3D model. <-- CHANGE
        mesh_simplify (float): The mesh simplification factor.
        texture_size (int): The texture resolution.
    Returns:
        str: The path to the extracted GLB file (for Model3D component).
        str: The path to the extracted GLB file (for DownloadButton).
    """
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)

    # --- Unpack state from the dictionary --- VITAL CHANGE for API
    gs, mesh = unpack_state(state_dict)

    # --- Postprocessing and Export ---
    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()
    # Return path twice for both Model3D and DownloadButton components
    return glb_path, glb_path


@spaces.GPU
def extract_gaussian(state_dict: dict, req: gr.Request) -> Tuple[str, str]: # <-- CHANGE: Accept dict
    """
    Extract a Gaussian file from the 3D model state dictionary.
    Args:
        state_dict (dict): The serializable dictionary state of the generated 3D model. <-- CHANGE
    Returns:
        str: The path to the extracted Gaussian file (for Model3D component).
        str: The path to the extracted Gaussian file (for DownloadButton).
    """
    user_dir = os.path.join(TMP_DIR, str(req.session_hash))
    os.makedirs(user_dir, exist_ok=True)

    # --- Unpack state from the dictionary --- VITAL CHANGE for API
    gs, _ = unpack_state(state_dict)

    gaussian_path = os.path.join(user_dir, 'sample.ply')
    gs.save_ply(gaussian_path)
    torch.cuda.empty_cache()
    # Return path twice for both Model3D and DownloadButton components
    return gaussian_path, gaussian_path


# --- Gradio UI Definition ---
# output_buf = gr.State() # No change needed here, it will now hold the dict
# video_output = gr.Video(...) # No change needed

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)

    # --- State Buffer ---
    # This will now hold the dictionary returned by text_to_3d
    output_buf = gr.State()

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

    # --- Generate Button Click Flow ---
    # No changes needed to the structure, but text_to_3d now puts the dictionary into output_buf
    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],
        outputs=[output_buf, video_output], # output_buf receives state_dict
    ).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 Button Click Flow ---
    # The input 'output_buf' now contains the state_dict needed by the modified extract_glb function
    extract_glb_btn.click(
        extract_glb,
        inputs=[output_buf, mesh_simplify, texture_size], # Pass the state_dict via output_buf
        outputs=[model_output, download_glb],
    ).then(
        lambda: gr.Button(interactive=True),
        outputs=[download_glb],
    )

    # --- Extract Gaussian Button Click Flow ---
    # The input 'output_buf' now contains the state_dict needed by the modified extract_gaussian function
    extract_gs_btn.click(
        extract_gaussian,
        inputs=[output_buf], # Pass the state_dict via output_buf
        outputs=[model_output, download_gs],
    ).then(
        lambda: gr.Button(interactive=True),
        outputs=[download_gs],
    )

    model_output.clear(
        lambda: gr.Button(interactive=False), # Should clear both potentially?
        outputs=[download_glb, download_gs], # Clear both download buttons
    )


# --- Launch the Gradio app ---
if __name__ == "__main__":
    # Consider adding error handling for pipeline loading
    try:
        pipeline = TrellisTextTo3DPipeline.from_pretrained("JeffreyXiang/TRELLIS-text-xlarge")
        # Move to GPU if available
        if torch.cuda.is_available():
            pipeline.cuda()
        else:
            print("WARNING: CUDA not available, running on CPU (will be very slow).")
        print("✅ Trellis pipeline loaded successfully.")
    except Exception as e:
        print(f"❌ Failed to load Trellis pipeline: {e}", file=sys.stderr)
        # Optionally exit if pipeline is critical
        # sys.exit(1)

    demo.launch()