ComfyUI-BerniniRWrapper

BerniniR_AttentionConfig

Attention backend selector. 'auto' = best available.

Bernini R/Sampling · 1 input · 1 parameter · 1 output

BerniniR_BlockSwapArgs

Configure block-swap VRAM offloading. Connect to Bernini-R KSampler

Bernini R/Config · 0 inputs · 6 parameters · 1 output

BerniniR_CLIPLoader

Load the Wan text encoder for Bernini-R prompt encoding.

Bernini R/Loaders · 0 inputs · 3 parameters · 1 output

BerniniR_CompileModel

Store torch.compile config on a Bernini-R model handle. Applied when the sampler loads the model.

Bernini R/Loaders · 2 inputs · 3 parameters · 1 output

BerniniR_Conditioning

Creates latent and attaches in-context visual references for Bernini-R.

Bernini R/Conditioning · 7 inputs · 8 parameters · 3 outputs

BerniniR_ContextWindow

Temporal context windows for long video generation.

Bernini R/Sampling · 2 inputs · 5 parameters · 1 output

BerniniR_DualExpertSampler

Dual-expert sampler: high_noise model for early steps, low_noise model for late steps. Block swap saves VRAM.

Bernini R/Sampling · 1 output

BerniniR_GuidanceConfig

Guidance strategy selector. Connect to BerniniR_KSampler to override

Bernini R/Config · 1 input · 3 parameters · 1 output

BerniniR_GuidanceStrengthSchedule

Pre-computed per-step guidance ceiling list for dynamic guidance.

Bernini R/Sampling · 1 input · 5 parameters · 1 output

BerniniR_KSampler

Enhanced KSampler with temporal context window tiling for long videos.

Bernini R/Sampling · 1 output

BerniniR_LoadLoRA

Queue a Wan/Bernini-R LoRA on a model handle.

Bernini R/Loaders · 1 input · 2 parameters · 1 output

BerniniR_ModelLoader

Create a lazy-loading Bernini-R model handle. The actual weights are loaded by the sampler when needed.

Bernini R/Loaders · 1 input · 1 parameter · 1 output

BerniniR_PromptEmbedding

Task-aware prompt planner: selects system prompt, encodes text with CLIP, caches embeddings.

Bernini R/Conditioning · 2 inputs · 9 parameters · 4 outputs

BerniniR_SegmentSchedule

Segment prompt-travel schedule: each segment has a single prompt

Bernini R/Conditioning · 1 input · 10 parameters · 2 outputs

BerniniR_TeaCacheArgs

Configure TeaCache block caching for faster sampling.

Bernini R/Config · 0 inputs · 5 parameters · 1 output

BerniniR_VAEDecode

Tiled VAE decode with optional colour matching.

Bernini R/VAE · 4 inputs · 6 parameters · 1 output

BerniniR_VAEEncode

Tiled VAE encode from image/video pixels to latent.

Bernini R/VAE · 2 inputs · 5 parameters · 1 output

BerniniR_VAELoader

Load the Bernini-R VAE for encoding/decoding video latents.

Bernini R/Loaders · 0 inputs · 1 parameter · 1 output