Instructions to use Viggle/Meridian with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Viggle/Meridian with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Viggle/Meridian", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Release checklist
Viggle/Meridian has been public since 2026-09-15. The 2026-09-18 update replaced the first
release's fused transformer with two LoRA adapters; the original weights remain under legacy/.
This page records what shipped and which checks are still open.
Settled
- Weights are uploaded.
teacher_lora/is the re-camera adapter on the unmodified MiniMax-H3 transformer;turbo_lora/is the DMD student distilled against it.legacy/transformer/andlegacy/lora/are the first release, kept so earlier results stay reproducible. - The Hub destination is
Viggle/Meridian. The olderViggle/Viggle-Recamidentifier was never a working download and must not be restored as one. - Uploads go through the release directory's own
push.py, which uses allow-patterns only. The archivedpush_docs.pytargets a different repository and is not a current uploader.
Still open
- Public-use clearance, including for the NBA footage and figures, remains pending. The presentation examples are research previews; retain source credits and edit records before choosing publication assets.
- Low-memory configurations, including RTX 4090, are community integration targets, not validated support.
- Fast point-cloud preview helps inspect an authored path; it does not guarantee generated-view fidelity.
Presentation and review
- Replace local preview links with cleared publication assets.
- Method video strips use actual matched frames; the 3D points and camera path are explicitly schematic. Retain this distinction and the source provenance.
- Studio overview is recorded and preview-only; capture and review a matching generated take before extending it to demonstrate final generation.
- Complete continuous visual motion review; automated playback and sampled frames are not sufficient.
- Retain exact camera controls if presenting an isolated space/time ablation.
- Confirm permission to use every source clip and to publish the corresponding demonstrations.
- Keep source-time labels accurate when comparing live, retimed, and held sequences.
- Verify the reported timing against a retained run log before publishing it as a headline result.
- Do not add a speedup ratio, quality comparison, ablation, or metric without supporting results.
- Upload referenced presentation media with the Markdown.