evoneural/evoneuralIn3D-app
0
Evoneural MVP – Mesh & Skybox
Text → 360° skybox (Stable Diffusion) and Text → 3D mesh (Stable Diffusion + TripoSR).
- Skybox: Enter a prompt → 2:1 equirectangular image.
- Mesh: Enter a prompt or upload an image → 3D mesh (.glb / .obj).
Required: Hugging Face token (this Space)
Skybox and mesh both need to download Stable Diffusion from the Hub on first run. Add a token so the Space can download the model:
- Go to huggingface.co/settings/tokens and create a token (read access is enough).
- In this Space: Settings → Variables and secrets → New secret.
- Name:
HF_TOKEN, Value: your token (e.g.hf_...). - Save. Restart the Space if it’s running.
After that, run Skybox or Mesh again; the model will download on first use.
Ready to use
- Push this repo to a Docker Space (SDK: docker, port 7860).
- Set the HF_TOKEN secret (Settings → Variables and secrets).
- Restart the Space. First build may take 15–25 minutes (installing PyTorch, TripoSR, torchmcubes).
- Open the app and generate a skybox or mesh. First run will download the Stable Diffusion model (~4 GB).
Troubleshooting
- "Could not load Stable Diffusion" → Add
HF_TOKENin Settings → Variables and secrets, then restart the Space. - "TripoSR not found" → Rebuild the Space (the Dockerfile clones TripoSR). If you forked the Space, ensure the Dockerfile and
scripts/are present. - Mesh generation times out → Use Mesh resolution 256 and disable Bake texture atlas for a faster run; GPU Spaces are much faster than CPU.
- Build fails on torchmcubes → The Dockerfile installs CPU-only PyTorch first (
--index-url https://download.pytorch.org/whl/cpu) so torchmcubes can build without CUDA (Docker build has no GPU). Do not change the order ofpip installsteps; do not remove the CPU-index torch install or the build will fail with "Your installed Caffe2 version uses CUDA but I cannot find CUDA". - "XOpenDisplay: cannot open display" / TripoSR baking texture fails → TripoSR’s texture baking uses ModernGL, which needs an X11 display. The image uses entrypoint.sh to start Xvfb in the background and then Streamlit so a virtual display is available. Do not remove
xvfbor change the ENTRYPOINT or baking will fail in the Space. - Launch timed out / workload not healthy → The app uses entrypoint.sh (Xvfb + Streamlit) and .streamlit/config.toml (headless) so startup is reliable. If it still times out, check Space logs for OOM or errors; consider upgrading to a larger machine.
Model: evoneural/evoneuralIn3D
