Model reference · open weights

Bernini-Wan2.2-scaled

Bernini-Wan2.2-scaled is an open-weight video model from attashe, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.

Video attashe 1 variants 8k downloads/mo
Request this model on EU hardware All served models Not on the shared API today — deployed on request.

About

What Bernini-Wan2.2-scaled is

Bernini (full) Wan2.2 renderer — fp8 e4m3 scaled (ComfyUI) The two DiT renderer transformers of the full ByteDance Bernini pipeline (diffdec = high-noise expert, diffdeclow = low-noise expert), quantized to fp8 e4m3 scaled in the ComfyUI format. The layout is byte-for-byte structurally identical to Comfy-Org/Bernini-R's wan2.2berninirfp8scaled.safetensors (verified: same 1815 keys, shapes, dtypes, and metadata) — the difference is only the weights, which here are the full Bernini renderer (jointly trained with the MLLM planner) rather than the renderer-only Bernini-R. Files Drop them into ComfyUI/models/diffusionmodels/ and use them anywhere the Bernini-R fp8scaled files work (same modeltype, same keys). Quantization details - Format marker per quantized weight: comfyquant = {"format": "float8e4m3fn"}. - Quantized Linears: selfattn.{q,k,v,o}, crossattn.{q,k,v} (cross-attn o kept in fp16), ffn.0, ffn.2 — 9 per block × 40 = 360 weights per expert. - For each quantized weight W: scale = max(|W|)/448, Wfp8 = (W/scale).clamp(±448).to(float8e4m3fn), stored alongside a scalar weightscale (fp32). Dequant: W ≈ Wfp8.to(dtype) weightscale. - Everything else (norms, modulation, patchembedding, text/timeembedding, timeprojection, head, all biases) is kept in fp16. - Mean per-tensor reconstruction error ≈ 2.2%. - Source: extracted from ByteDance/Bernini-Diffusers (bernini/ checkpoint, fp32), with diffusers WanTransformer3DModel keys remapped to the original Wan / ComfyUI naming. License: Apache-2.0, inherited from the upstream Bernini release.

Summarised from the published model card. Read the full card on the HuggingFace links below.

Specifications

What it is

Makerattashe
TypeVideo models
Variants1
Runs withdiffusers
Based onByteDance/Bernini-Diffusers
Released2026-06-16
Popularity8k downloads / month
Likes2
LicenceOpen weights

How it works

How video models work

Prompt / imagestart pointTemporal diffusionframes over timeVideoMP4 clipA video model generates a sequence of coherent frames from your prompt or a starting image.

Variants

Sizes & precisions

Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.

VariantParamsPrecisionVRAMFits 16 GBWeights
Bernini-Wan2.2-fp8-scaledFP8Weights ↗

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys bernini-wan2-2-scaled for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (bernini-wan2-2-scaled below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/videos/generations \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"bernini-wan2-2-scaled","prompt":"a drone shot over a forest"}'

Details

Languages, data & research

Tags

diffusers safetensors wan2.2 bernini fp8 comfyui text-to-video

Licence

Open weights

Open weights under apache-2.0 — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗

Sources

Weights & code

Want Bernini-Wan2.2-scaled on EU-owned hardware?

Request this model on EU hardware See what’s served now

Explore

More video models

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms