---
title: Bernini-v2-ComfyUI
canonical_url: "https://www.modelscope.cn/models/rzgar/Bernini-v2-ComfyUI"
md_url: "https://www.modelscope.cn/models/rzgar/Bernini-v2-ComfyUI.md"
repository: rzgar/Bernini-v2-ComfyUI
last_updated: 2026-08-31
license: apache-2.0
model_type:
  - bernini
architectures:
  - BerniniModel
base_model:
  - ByteDance/Bernini-Diffusers-v2
base_model_relation: finetune
parameters: 101.1B
tensor_type:
  - F16
  - BF16
  - F8_E4M3
  - U8
  - F32
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 168
stars: 0
tags:
  - text-to-video
  - image-to-video
  - video-to-video
  - image-to-image
  - comfyui
---

# Bernini-v2-ComfyUI

> Bernini-v2-ComfyUI - rzgar 在 ModelScope 开源的模型。Bernini v2 is a video editor: t2v / i2v / v2v / r2v / rv2v / ads2v. The source media (reference images / source video) are first-class inputs. The pipeline has two stages:

rzgar/Bernini-v2-ComfyUI 是 ModelScope 魔搭社区上的 101.1B 参数机器学习模型，采用 apache-2.0 许可，基于 ByteDance/Bernini-Diffusers-v2 构建。

- **Repository**: rzgar/Bernini-v2-ComfyUI
- **License**: apache-2.0
- **Parameters**: 101.1B
- **Base model**: ByteDance/Bernini-Diffusers-v2
- **Tags**: text-to-video, image-to-video, video-to-video, image-to-image, comfyui
- **Downloads**: 168
- **Stars**: 0
- **Last updated**: 2026-08-31

Source: https://www.modelscope.cn/models/rzgar/Bernini-v2-ComfyUI

---

Bernini v2 is a video *editor*: t2v / i2v / v2v / r2v / rv2v / ads2v. The source media (reference images / source video) are first-class inputs. The pipeline has two stages:

1. **Semantic planning**: a fine-tuned Qwen2.5-VL (`mllm`) + connector + `vit_decoder` (MaskGIT-style loop) turn the text + source visual tokens into `diff_mllm_contexts` (the novel part Bernini-R lacks).
2. **Rendering**: two co-trained Wan2.2 DiTs (**high-noise** / **low-noise**, switched at 0.875) with the Bernini `vae_txt_vit_wapg` guidance (4-way CFG with APG projection), VAE-encoded source media as `context_latents`, Wan2.1 VAE decode.

## ComfyUI Custom node
Extract the [ZIP](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/ComfyUI-BerniniV2/ComfyUI-BerniniV2.zip) file in `ComfyUI/custom_nodes`
<img src="https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/workflow/custom_node_screenshot.png" width="1280" height="720" />

Task is inferred from which media inputs are connected: (none) = t2v, `source_video` = v2v, `source_video` + `reference_images` = rv2v, `reference_images` only = r2v, `source_video` + `reference_video` = ads2v.

**Recommended settings**
- CFG: 3
- Steps: 16, 23 or 50
- Sampler: UniPC / DPMPP_2M / EULER 



## Model files
- [FP8_scaled](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/tree/main/Bernini_v2_FP8_scaled)

| Component |File |
|---|---|
| high-noise DiT | [wan2.2_bernini_v2_high_noise_nvfp4.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/Bernini_v2_NVFP4/wan2.2_bernini_v2_high_noise_nvfp4.safetensors) |
| low-noise DiT | [wan2.2_bernini_v2_low_noise_nvfp4.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/Bernini_v2_NVFP4/wan2.2_bernini_v2_low_noise_nvfp4.safetensors) |
| T5 (UMT5-XXL) | [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/text_encoders) |
| VAE (Wan2.1) |  [Wan2_1_VAE_fp32.safetensors](https://huggingface.co/rzgar/Qwen-Image-WAN2.2-VAE-Collection/blob/main/Vanilla/Wan2_1_VAE_fp32.safetensors) |

Place the planner files **anywhere under `models/text_encoders/`** - _folder names don't matters_:

| Component | File | Where |
|---|---|---|
| planner MLLM| [bernini_mllm_fp8_scaled.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/text_encoders/Bernini/bernini_mllm_fp8_scaled.safetensors) | `text_encoders/Bernini/` |
| connector | [connector.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/text_encoders/Bernini/connector.safetensors) | text_encoders/Bernini/ |
| vit_decoder | [vit_decoder_bf16.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/text_encoders/Bernini/vit_decoder_bf16.safetensors) | text_encoders/Bernini/ |
| mask_tokens| [mask_tokens.safetensors](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/text_encoders/Bernini/mask_tokens.safetensors) | text_encoders/Bernini/ |
| planner tokenizer + processor configs |  [mllm folder](https://huggingface.co/rzgar/Bernini-v2-ComfyUI/resolve/main/text_encoders/Bernini/mllm/mllm.zip) | text_encoders/Bernini/mllm |



___