Functional Interface to Diffusion Models in R

---

Documentation for package ‘diffuseR’ version 0.2.2

• DESCRIPTION file.
• User guides, package vignettes and other documentation.
• Package NEWS.

Help Pages

A B C D E F G I J L M N O P Q R S T U V W Z

-- A --

audio_encode_ltx23	Audio Conditioning Frontend for LTX-2.3
audio_vae_ltx23	LTX-2.3 Audio VAE
auto_devices	Auto-Configure Device Assignment

-- B --

bpe_tokenizer

BPE Tokenizer

-- C --

checkpoint_flux	FLUX Checkpoint Readers
checkpoint_ltx23	LTX-2.3 Single-File Checkpoint Reader
clear_vram	Clear VRAM Cache
CLIPTokenizer	Tokenize a prompt
clip_pooled_output	Pooled CLIP output at the EOS position
condition_ltx23	LTX-2.3 Prefix Conditioning (image-to-video, video continuation)
connectors_ltx23	LTX-2.3 Text Embedding Connectors
convert_sd21_pt_to_diffusers	Convert cornball SD 2.1 TorchScript weights to a diffusers artifact

-- D --

ddim_scheduler_create	Create a DDIM Scheduler
ddim_scheduler_step	Perform a DDIM scheduler step
decode_bpe	Decode token IDs to text
dit_flux	FLUX Transformer (MMDiT)
dit_flux2	FLUX.2 Transformer (MMDiT)
dit_flux2_modules	FLUX.2 Transformer Building Blocks
dit_flux_modules	FLUX Transformer Building Blocks
dit_ltx23	LTX-2.3 Audio-Video Diffusion Transformer
dit_ltx23_modules	LTX-2.3 Transformer Building Blocks
dit_zimage_modules	Z-Image Transformer Block Modules
download_component	Download a single TorchScript model component
download_flux	Download and Prepare FLUX.1-schnell Weights
download_flux1	Download FLUX.1-schnell and build the quantized artifact
download_flux2	Download and Prepare FLUX.2 Klein 4B Weights
download_flux2_klein	Download FLUX.2-klein-4B and build the quantized artifact
download_ltx2	Download the LTX-2.3 checkpoint and build a quantized artifact
download_ltx23	Download and Prepare LTX-2.3 Model Weights
download_model	Download TorchScript model files for Stable Diffusion
download_sd21	Download the Stable Diffusion 2.1 diffusers weights
download_sdxl	Download the Stable Diffusion XL diffusers weights
download_zimage	Download and Prepare Z-Image-Turbo Weights
download_zimage_turbo	Download Z-Image-Turbo and build the quantized artifact

-- E --

encode_bpe	Encode text to token IDs
encode_qwen	Encode prompts with the Qwen tokenizer
encode_unigram	Encode text with a Unigram tokenizer
encode_with_gemma3	Encode text with Gemma3 for LTX-2
encode_with_qwen3	Encode prompts with the Qwen3 encoder for FLUX.2
encode_with_t5	Encode prompts with the T5 encoder

-- F --

filename_from_prompt	Generate a filename from a prompt
flowmatch_calculate_shift	Calculate shift for dynamic shifting
flowmatch_scale_noise	Scale noise for flow matching forward process
flowmatch_scheduler_create	Create a FlowMatch Euler Discrete Scheduler
flowmatch_scheduler_step	Perform a FlowMatch scheduler step
flowmatch_set_timesteps	Set timesteps for inference
flux2_bn_normalize	Normalize patchified latents with the VAE BatchNorm statistics
flux2_double_block	FLUX.2 double-stream (MMDiT) block
flux2_empirical_mu	Empirical timestep shift for FLUX.2
flux2_feed_forward	FLUX.2 feed-forward (fused SwiGLU)
flux2_is_quant_key	Test whether a FLUX.2 key is in the quantization cast set
flux2_load_pipeline	Load the FLUX.2 klein pipeline
flux2_modulation	FLUX.2 shared modulation projection
flux2_pack_latents	Pack patchified FLUX.2 latents into tokens
flux2_parallel_self_attention	FLUX.2 parallel self-attention (single-stream)
flux2_patchify_latents	Patchify FLUX.2 latents (2x2 -> channels)
flux2_prepare_latent_ids	Build FLUX.2 latent position ids
flux2_prepare_text_ids	Build FLUX.2 text position ids
flux2_single_block	FLUX.2 single-stream block (parallel)
flux2_transformer	FLUX.2 transformer model
flux2_unpack_latents_with_ids	Unpack FLUX.2 tokens back to a latent grid via position ids
flux2_unpatchify_latents	Unpatchify FLUX.2 latents (channels -> 2x2)
flux2_vae_decoder	FLUX.2 VAE decoder
flux_ada_layer_norm_continuous	FLUX continuous adaLN (final norm)
flux_ada_layer_norm_zero	FLUX adaLN-Zero modulation (double-stream)
flux_ada_layer_norm_zero_single	FLUX adaLN-Zero modulation (single-stream)
flux_apply_rotary_emb	Apply FLUX rotary embeddings to a per-head tensor
flux_attention	FLUX joint attention
flux_double_block	FLUX double-stream (MMDiT) transformer block
flux_is_quant_key	Test whether a FLUX key is in the quantization cast set
flux_load_pipeline	Load the FLUX.1-schnell pipeline
flux_load_transformer	Load a FLUX transformer from any checkpoint format
flux_memory_profile	Resolve a FLUX memory profile
flux_open_checkpoint	Open a FLUX transformer checkpoint directory
flux_open_quantized	Open a quantized FLUX artifact directory
flux_pack_latents	Pack FLUX latents into a patch sequence
flux_pos_embed	Compute FLUX rotary frequencies from position ids
flux_prepare_latent_image_ids	Build FLUX latent image position ids
flux_quantize	Quantize a FLUX transformer to NF4 or fp8 shards
flux_single_block	FLUX single-stream transformer block
flux_transformer	FLUX transformer model
flux_unpack_latents	Unpack a FLUX patch sequence back into latents
fp8_ltx23	FP8 Weight Storage for the LTX-2.3 Transformer

-- G --

gemma3_config_ltx2	Create Gemma3 configuration for LTX-2
gemma3_encode_batch	Batch-encode prompts with Gemma3, cached to disk
gemma3_quantize_nf4	Quantize a Gemma3 text encoder to NF4 shards
gemma3_text_model	Gemma3 Text Model
gemma3_tokenizer	Gemma3 Tokenizer
get_required_components	Get required components for each model type

-- I --

img2img	Image-to-Image Generation with Stable Diffusion
is_blackwell_gpu	Check if GPU is Blackwell Architecture

-- J --

jit_ltx23	LTX-2.3 JIT Block Stack
jit_vae_ltx23	JIT-Traced Decode for the LTX-2.3 VAEs and Vocoder

-- L --

latents_to_video	Create Video from Latents (Helper)
load_decoder_safetensors	Load HF safetensors VAE weights into the native decoder
load_decoder_weights	Load weights from TorchScript decoder into native decoder
load_flux2_vae_decoder	Load the FLUX.2 VAE decoder from safetensors
load_gemma3_nf4	Load a Gemma3 text encoder from an NF4 artifact
load_gemma3_text_encoder	Load Gemma3 Text Model from safetensors
load_model_component	Load a specific component of a diffusion model
load_pipeline	Load a diffusion model pipeline
load_qwen3_text_encoder	Load a Qwen3 encoder from a transformers directory
load_t5_text_encoder	Load a T5 encoder from a transformers directory
load_text_encoder2_safetensors	Load HF safetensors weights into the native SDXL text encoder 2
load_text_encoder2_weights	Load weights from TorchScript text encoder 2 into native encoder
load_text_encoder_safetensors	Load HF safetensors weights into the native CLIP text encoder
load_text_encoder_weights	Load weights from TorchScript text encoder into native encoder
load_to_gpu	Load Module to GPU
load_unet_safetensors	Load HF safetensors weights into the native SD21 UNet
load_unet_sdxl_safetensors	Load HF safetensors weights into the native SDXL UNet
load_unet_sdxl_weights	Load weights from TorchScript SDXL UNet into native SDXL UNet
load_unet_weights	Load weights from TorchScript UNet into native UNet
ltx23_adain_filter_latent	Adaptive instance normalization between latent tensors
ltx23_ada_layer_norm_single	Adaptive layer norm single (adaLN-single)
ltx23_antialias_act1d	Anti-aliased activation
ltx23_apply_interleaved_rotary_emb	Apply interleaved rotary embeddings
ltx23_apply_split_rotary_emb	Apply split rotary embeddings
ltx23_attention	LTX-2 attention layer
ltx23_audio_causal_conv2d	Causal 2D convolution for audio spectrograms
ltx23_audio_decoder	LTX-2.3 audio VAE decoder
ltx23_audio_downsample	LTX audio downsampler
ltx23_audio_encoder	LTX-2.3 audio VAE encoder
ltx23_audio_mel_frontend	Build the 16 kHz log-mel frontend for audio conditioning
ltx23_audio_resnet_block	LTX audio ResNet block
ltx23_audio_upsample	LTX audio upsampler
ltx23_audio_vae	LTX-2.3 audio VAE
ltx23_causal_conv3d	Causal 3D convolution
ltx23_census	Summarize checkpoint key coverage
ltx23_connector_transformer_1d	1D connector transformer
ltx23_denormalize_latents	Denormalize latents with the VAE's per-channel statistics
ltx23_distilled_sigmas	Official distilled sigma schedule
ltx23_downsample1d	Anti-aliasing 1D downsampler (low-pass then stride)
ltx23_encode_audio	Encode audio into normalized, packed conditioning latents
ltx23_encode_video_frames	Encode pixel frames to normalized video latents
ltx23_feed_forward	LTX feed-forward layer
ltx23_fp8_linear	FP8 linear layer
ltx23_get_timestep_embedding	Sinusoidal timestep embedding
ltx23_is_fp8_cast_key	Test whether a mapped DiT key is in the official fp8 cast set
ltx23_kaiser_sinc_filter1d	Kaiser sinc low-pass filter kernel
ltx23_latent_upsampler	LTX-2.3 latent upsampler model
ltx23_load_group	Stream a checkpoint key group into a module
ltx23_load_pipeline	Load the LTX-2.3 generation components from a single-file checkpoint
ltx23_load_transformer_fp8	Load the LTX-2.3 transformer with FP8 weights
ltx23_load_transformer_nf4	Load the LTX-2.3 transformer with resident NF4 weights
ltx23_load_upsampler	Load the LTX-2.3 spatial upscaler weights
ltx23_map_audio_vae_key	Map an official audio VAE checkpoint key to the R module name
ltx23_map_connector_key	Map an official connectors checkpoint key to the R module name
ltx23_map_dit_key	Map an official DiT checkpoint key to the R module name
ltx23_map_vae_key	Map an official VAE checkpoint key to the R module name
ltx23_map_vocoder_key	Map an official vocoder checkpoint key to the R module name
ltx23_mel_stft	Causal log-mel spectrogram with checkpoint-loaded bases
ltx23_memory_profile	Get an LTX-2.3 memory profile
ltx23_nf4_dequantize	Dequantize NF4 data to a float tensor
ltx23_nf4_linear	NF4 linear layer
ltx23_nf4_quantize	Quantize a tensor to NF4
ltx23_normalize_latents	Normalize latents with the VAE's per-channel statistics
ltx23_open_checkpoint	Open an LTX-2.3 checkpoint
ltx23_open_fp8_checkpoint	Open an FP8 shard directory as a checkpoint
ltx23_per_channel_rms_norm	Per-channel RMS normalization
ltx23_per_token_rms_norm	Per-token RMS normalization over the channel axis
ltx23_prepare_conditioned_latents	Build conditioned initial latents and the conditioning mask
ltx23_preprocess_frames	Preprocess an image (or frame stack) for VAE encoding
ltx23_quantize_fp8	Quantize an LTX-2.3 checkpoint to FP8 shards
ltx23_quantize_nf4	Quantize an LTX-2.3 checkpoint to NF4 shards
ltx23_read_audio	Read an audio file as 16 kHz stereo PCM
ltx23_read_tail_frames	Read the trailing frames of a video file
ltx23_release_dequant_buffers	Release the NF4 dequantization buffers
ltx23_rms_norm	RMS normalization
ltx23_rotary_pos_embed	LTX-2.3 audio/video rotary position embedder
ltx23_rotary_pos_embed_1d	1D rotary embeddings for the text connectors
ltx23_set_attn_chunk	Set the attention query-chunk size across a transformer
ltx23_snake_beta	SnakeBeta activation
ltx23_split_keys	Split checkpoint keys by component
ltx23_stage2_distilled_sigmas	Stage-2 distilled sigma schedule (two-stage refinement)
ltx23_tail_latents	Slice the trailing latent frames of a generation for chaining
ltx23_text_connectors	LTX-2.3 text connectors
ltx23_tone_map_latents	Sigmoid tone mapping for latents
ltx23_transformer	LTX-2.3 video transformer model
ltx23_transformer_block	LTX-2 transformer block
ltx23_tune_gc	Tune the torch CUDA allocator for large-resident inference
ltx23_upsample1d	Anti-aliasing 1D upsampler (transposed low-pass)
ltx23_video_decoder3d	LTX-2.3 video decoder
ltx23_video_downsampler3d	Pixel-unshuffle 3D downsampler
ltx23_video_down_block3d	LTX video down block
ltx23_video_encoder3d	LTX-2.3 video encoder
ltx23_video_mid_block3d	LTX video mid block
ltx23_video_resnet_block3d	LTX 3D ResNet block
ltx23_video_upsampler3d	Pixel-shuffle 3D upsampler
ltx23_video_up_block3d	LTX video up block
ltx23_video_vae	LTX-2.3 video VAE
ltx23_vocoder	LTX-2.3 vocoder stage
ltx23_vocoder_resblock	Vocoder ResNet block (AMP)
ltx23_vocoder_with_bwe	LTX-2.3 vocoder with bandwidth extension

-- M --

memory_flux	FLUX Memory Profiles
memory_ltx23	LTX-2.3 Memory Profiles and CUDA GC Tuning
models2devices	models2devices

-- N --

nf4_ltx23

NF4 Weight Storage for the LTX-2.3 Transformer

-- O --

offload_to_cpu

Offload Module to CPU

-- P --

post_quant_conv	Post Quant Conv
preprocess_image	Preprocess image for Stable Diffusion
print.bpe_tokenizer	Print BPE Tokenizer
print.diffuseR_resident	Print a resident handle

-- Q --

quantize_flux	FLUX Transformer Quantization and Loading
quant_conv	Quant Conv
qwen3_encoder	Qwen3 encoder stack
qwen3_text_encoder	Qwen3 Text Encoder
qwen_bpe_tokenizer	Load a Qwen2 byte-level BPE tokenizer

-- R --

recommend	Recommend a precision and device configuration for a model
reshard_safetensors	Re-shard a large safetensors file into sub-2 GB shards
resident_activate	Bring a resident pipeline onto the GPU
resident_deactivate	Release a resident pipeline's VRAM
resident_generate	Generate from an active resident pipeline
resident_load	Load a diffusion pipeline as a resident handle
resident_status	Status of a resident handle
resident_unload	Drop a resident handle entirely
rope_flux	FLUX Rotary Positional Embeddings
rope_flux2	FLUX.2 Position Ids and Empirical Shift
rope_ltx23	LTX-2.3 Rotary Positional Embeddings
rope_zimage	Z-Image Rotary Positional Embeddings and Patchify Helpers

-- S --

save_frames	Save Video Frames as Individual Images
save_image	Save and Display an Image from a Torch Tensor
save_video	Save Video to File
save_video_ltx23	Save an LTX video (optionally with audio) to MP4
scheduler_add_noise	Add noise to latents using DDIM scheduler
sdxl_memory_profile	Get SDXL Memory Profile
sdxl_pipeline_from_safetensors	Assemble a native SDXL pipeline from a diffusers safetensors directory
sdxl_pipeline_safetensors	Native SDXL pipeline from diffusers safetensors
sd_pipeline_from_safetensors	Assemble a native SD pipeline from a diffusers safetensors directory
sd_pipeline_safetensors	Native Stable Diffusion pipelines from diffusers safetensors
serve	Serve diffuseR over HTTP
setup_dtype	Set up dtype based on device configuration
staging	Pinned Staging for Phase-Sequential Components
staging_ltx23	Pinned Staging for Phase-Sequential Components
standardize_devices	Standardize devices configuration
st_caps	safetensors read-capability probes and fork messaging

-- T --

t5_encoder	T5 encoder stack
t5_text_encoder	T5 Text Encoder (T5-v1.1)
text_encoder2_native	Native CLIP Text Encoder 2 (OpenCLIP ViT-bigG for SDXL)
text_encoder2_native_from_safetensors	Build a native SDXL text encoder 2 from a diffusers safetensors directory
text_encoder_native	Native CLIP Text Encoder
text_encoder_native_from_safetensors	Build a native CLIP text encoder from a diffusers safetensors directory
tokenizer_qwen	Qwen2 Byte-Level BPE Tokenizer
tokenizer_unigram	SentencePiece Unigram Tokenizer
tokenize_gemma3	Tokenize text for Gemma3
txt2img	Generate an image from a text prompt using a diffusion pipeline
txt2img_flux	Generate an image with FLUX.1-schnell
txt2img_flux2	Generate an image with FLUX.2 klein
txt2img_sd21	Generate an image from a text prompt using a diffusion pipeline
txt2img_sdxl	Generate an image from a text prompt using SDXL
txt2img_zimage	Generate an image with Z-Image-Turbo
txt2vid_ltx2	Generate video (and audio) with LTX-2.3
txt2vid_ltx23	LTX-2.3 Text-to-Video Pipeline

-- U --

unet_native	Native UNet for Stable Diffusion
unet_native_from_safetensors	Build a native SD21 UNet from a diffusers safetensors directory
unet_native_from_torchscript	Create native UNet from TorchScript
unet_safetensors	Load HF safetensors weights into the native SD/SDXL UNet
unet_sdxl_native	Native SDXL UNet
unet_sdxl_native_from_safetensors	Build a native SDXL UNet from a diffusers safetensors directory
unet_sdxl_native_from_torchscript	Create native SDXL UNet from TorchScript
unigram_tokenizer	Load a Unigram tokenizer from tokenizer.json
upsampler_ltx23	LTX-2.3 Spatial Latent Upsampler

-- V --

vae_decoder_native	Native VAE Decoder
vae_decoder_native_from_safetensors	Build a native VAE decoder from a diffusers safetensors directory
vae_flux2	FLUX.2 Latent Layout and VAE Helpers
vae_ltx23	LTX-2.3 Causal Video VAE
vae_ltx23_modules	LTX-2.3 Video VAE Building Blocks
vocab_size	Get vocabulary size
vocoder_ltx23	LTX-2.3 Vocoder with Bandwidth Extension
vram	VRAM Detection and Management Utilities
vram_report	Report VRAM Usage

-- W --

write_wav

Write a 16-bit PCM WAV file

-- Z --

zimage_block	Z-Image transformer block
zimage_cap_pos_ids	Build Z-Image caption position ids
zimage_feed_forward	Z-Image feed-forward (SwiGLU with separate gate weights)
zimage_final_layer	Z-Image final layer
zimage_img_pos_ids	Build Z-Image latent image position ids
zimage_is_quant_key	Test whether a Z-Image key is in the quantization cast set
zimage_load_pipeline	Load the Z-Image-Turbo pipeline
zimage_patchify	Patchify a latent image to Z-Image tokens
zimage_pos_embed	Compute Z-Image rotary frequencies from position ids
zimage_transformer	Z-Image Transformer
zimage_t_embedder	Z-Image timestep embedder
zimage_unpatchify	Unpatchify Z-Image tokens back to a latent image