Skip to content

NODESREADME.md

Latest commit

 

History

History
743 lines (494 loc) · 23.1 KB

NODESREADME.md

File metadata and controls

743 lines (494 loc) · 23.1 KB

NODESREADME — Boyonodes Detailed Reference

Full node-by-node documentation. For a one-line summary of every node see README.md.

Audio & TTS

Boyo Chatterbox Turbo Loader

File: boyo_chatterbox_turbo_loader.py | Category: Boyo/Audio/TTS

Loads the Chatterbox Turbo text-to-speech model into memory. Run this once and feed the model handle into the generate node.

Inputs

  • device — cuda or cpu

Outputs

  • model — Chatterbox model handle

Notes: Install with pip install chatterbox-tts --no-deps and pip install resemble-perth --no-deps. The --no-deps flags are required to avoid conflicts with ComfyUI's own dependency stack.

Boyo Chatterbox Turbo Generate

File: boyo_chatterbox_turbo_generate.py | Category: Boyo/Audio/TTS

Generate speech from text. Supports emotion tags inline in the transcript and optional voice cloning from a reference audio clip.

Inputs

  • model — handle from Loader node
  • text — transcript; embed emotion tags anywhere e.g. Hello [chuckle] how are you?
  • reference_audio (optional) — 5+ second audio clip to clone the voice from
  • exaggeration — emotion intensity multiplier
  • cfg_weight — classifier-free guidance weight
  • seed

Outputs

  • audio — ComfyUI AUDIO tensor at 24 kHz

Supported emotion tags: [clear throat] [sigh] [shush] [cough] [groan] [sniff] [gasp] [chuckle] [laugh]

Boyo Audio Duration Analyzer

File: boyo_audio_duration_analyzer.py | Category: Boyo/Audio/Analysis

Extracts the precise duration from a ComfyUI audio tensor. Handles both (batch, channels, samples) and (batch, samples, channels) layouts automatically.

Inputs

  • audio — ComfyUI AUDIO tensor

Outputs

  • duration_seconds FLOAT
  • sample_rate INT
  • info STRING — human-readable summary

Boyo Audio Padder

File: boyo_audio_padder.py | Category: Boyo/Audio/Processing

Pads audio with silence to match a target duration. Solves the classic ComfyUI lip-sync timing problem where generated audio is shorter than the video clip.

Inputs

  • audio — source audio tensor
  • target_duration FLOAT — desired total length in seconds
  • auto_center BOOLEAN — pad equally before and after instead of only at the end
  • sample_rate INT — output sample rate (default 12 kHz for correct ComfyUI playback speed)

Outputs

  • audio — padded audio tensor
  • info STRING — e.g. 📊 Audio: 3.2s + Padding: 7.3s = Total: 10.5s ✅ Perfect match!

Typical workflow: TTS Generate → Boyo Audio Padder (target = video length) → VHS Save Audio

Boyo Voice Enhancer

File: boyo_voice_enhancer.py | Category: Boyo/Audio/TTS

Enhances or converts voice characteristics using the Seed-VC model. Requires the seed-vc submodule and its dependencies.

Inputs

  • audio — source audio tensor
  • reference_audio — target voice reference
  • Various Seed-VC inference parameters (diffusion steps, pitch shift, etc.)

Outputs

  • audio — enhanced/converted audio tensor

Setup: git submodule add https://github.com/Plachtaa/seed-vc.git seed-vc && git submodule update --init --recursive

Boyo Audio Eval

File: BoyoAudioEval.py | Category: VideoUtils

Reads an audio file from disk and converts its duration to a frame count. Useful for sizing video generation to match audio length before any audio tensor is loaded into the workflow.

Inputs

  • audio_path STRING — absolute or relative path to a WAV file
  • fps FLOAT — target frames per second

Outputs

  • frame_count INT
  • video_length_seconds FLOAT
  • metadata STRING — human-readable summary

Image

Boyo Image Grab

File: boyo_image_grab.py | Category: Boyo/Image/Editing

Monitors a directory and loads the most recently modified image. Designed for iterative semantic editing chains where the output of one pass becomes the input of the next automatically.

Inputs

  • directory STRING
  • auto_refresh BOOLEAN
  • file_extension — filter by type

Outputs

  • image IMAGE
  • filename STRING

Boyo Paired Image Saver

File: boyo_paired_image_saver.py | Category: Boyo/Image/Editing

Saves an original and an edited image as a sequentially numbered pair. Compatible with ControlNet dataset formats. Use alongside Boyo Image Grab to build an iterative editing pipeline.

Inputs

  • original_image IMAGE
  • edited_image IMAGE
  • output_directory STRING
  • filename_prefix STRING

Outputs: none (output node)

Boyo Image Crop

File: boyo_image_crop.py | Category: Boyo

Tile-crops a large image into overlapping patches and saves them to a directory. Useful for building tiled datasets from high-resolution source material.

Inputs

  • image IMAGE
  • crop_width / crop_height INT — patch size
  • overlap INT — pixel overlap between adjacent patches
  • output_path STRING — destination directory

Outputs

  • status STRING — summary of patches written

Boyo Qwen VL Grounding

File: boyo_qwen_grounding.py | Category: (Boyo/Vision implied)

Runs Qwen2.5-VL visual grounding to locate objects in an image and draw annotated bounding boxes. Supports 4-bit quantisation for lower VRAM usage.

Inputs

  • image IMAGE
  • model_checkpoint — Qwen2.5-VL model path
  • prompt STRING — grounding query e.g. "the red car"
  • quantize_4bit BOOLEAN

Outputs

  • image IMAGE — annotated with bounding boxes
  • grounding_json STRING — raw detection results

Dependency: pip install qwen-vl-utils

LoRA Management

Boyo LoRA JSON Builder

File: boyo_lora_json_builder.py | Category: Boyo/LoRA

Creates and saves LoRA configuration JSON files. Each config can hold a paired high-noise and low-noise LoRA path plus an arbitrary list of trigger prompts. Utility LoRAs (e.g. lightning/turbo) can omit prompts entirely.

Inputs

  • config_name STRING
  • high_noise_lora STRING — filename from your loras folder
  • low_noise_lora STRING (optional)
  • prompts STRING — one prompt per line
  • save_directory STRING

Outputs

  • config_path STRING — path to the written JSON file

Boyo LoRA Paired Loader

File: boyo_lora_paired_loader.py | Category: Boyo/LoRA

Loads up to three LoRA configs at once and resolves prompts from each according to a per-slot strategy. Outputs all six LoRA paths and four combined prompt strings ready for downstream nodes.

Inputs

  • config_1/2/3 — config file names (dropdowns from your config directory)
  • strategy_1/2/3 — Mute, Concatenate, or Merge
  • prompt_mode — First Only, Cycle Through, or Random
  • seed INT — used for deterministic random prompt selection

Outputs

  • lora_high_1–3 STRING — high-noise LoRA paths
  • lora_low_1–3 STRING — low-noise LoRA paths
  • prompt_1–4 STRING — resolved combined prompts

Typical usage: slot 1 = utility LoRA (strategy: Mute), slots 2–3 = character + style (strategy: Concatenate).

Boyo LoRA Config Inspector

File: boyo_lora_config_inspector.py | Category: Boyo/LoRA

Reads a config file and prints a detailed breakdown: paired vs single LoRA type, all prompt strings, file existence checks, and usage recommendations. Connect before loading to debug config issues.

Inputs

  • config_file STRING

Outputs

  • report STRING — formatted inspection report

Boyo LoRA Config Processor

File: boyo_lora_config_processor.py | Category: Boyo/LoRA

The logic half of a split architecture. Reads a config and applies the chosen prompt mode and strategy, outputting resolved paths and prompt text. Pair with Boyo LoRA Path Forwarder.

Inputs

  • prompt_mode — First Only, Cycle Through, Random
  • config_1/2/3 (optional)
  • strategy_1/2/3
  • prepend_text / append_text STRING
  • seed INT

Outputs

  • Resolved LoRA paths and prompt strings

Boyo LoRA Path Forwarder

File: boyo_lora_path_forwarder.py | Category: Boyo/LoRA

Buffers LoRA filenames and forwards them to standard ComfyUI LoRA loader nodes. Caches the LoRA list at startup to prevent dynamic type-change issues during long batch runs.

Inputs

  • lora_high / lora_low STRING — filenames from Processor or Paired Loader

Outputs

  • lora_high / lora_low STRING — same filenames, now stable for downstream loaders

Boyo LoRA Info Sender

File: boyo_lora_info_sender.py | Category: Boyo/LoRA Tools

Minimal LoRA selector. Presents a dropdown of all available LoRAs and outputs just the selected filename as a string. Designed to feed index-switcher or routing nodes.

Inputs

  • lora_name — dropdown from loras folder

Outputs

  • lora_filename STRING

Boyo FramePack LoRA Loader

File: nodes.py | Category: BoyoNodes

Applies a LoRA to a FramePack (Hunyuan Video) model. Accepts a direct file path rather than a dropdown, standardises key format, and returns the patched model.

Inputs

  • model FramePackMODEL
  • lora_path STRING — absolute path to .safetensors
  • lora_strength FLOAT (0–2, default 1.0)

Outputs

  • model FramePackMODEL

Storyboard

Boyo Storyboard Prompt

File: boyo_storyboard_prompt.py | Category: Boyo/Storyboard

Constructs a system prompt and user prompt that instructs an ollama model to generate a structured 6-scene storyboard in JSON. Two modes are available: standard (6 image + 6 video prompts) and travelling (6 images + 6 multi-line video sequences for extended content).

Inputs

  • story_outline STRING
  • character_description STRING
  • style_description STRING
  • mode — Standard or Travelling
  • additional_details STRING (optional)
  • model_trigger_word STRING — LoRA/video model trigger

Outputs

  • system_prompt STRING
  • user_prompt STRING

Recommended model: Qwen 30B A3B Coder Abliterated (via ollama). Avoid Gemma, Meta coding variants, and thinking models.

Boyo Storyboard Output

File: boyo_storyboard_output.py | Category: Boyo/Storyboard

Parses the JSON string returned by an ollama node and splits it into 12 individual prompt outputs for direct connection to image and video generation nodes.

Inputs

  • json_string STRING — raw JSON from ollama

Outputs

  • image_1 … image_6 STRING
  • video_1 … video_6 STRING

Boyo Storyboard JSON Parser

File: boyo_storyboard_json_parser.py | Category: Boyo/Storyboard

Alternative parser for storyboard JSON with the same 12-output structure. Use if the primary output node has trouble with a particular model's JSON formatting.

Inputs / Outputs: identical to Boyo Storyboard Output.

Video

Boyo Video Clipper

File: boyo_video_clipper.py | Category: Boyo/Video

Clips a video from ComfyUI's input directory to an exact frame count starting from a given time. Uses FFmpeg internally. Useful for preparing dataset clips at a precise resolution and frame rate.

Inputs

  • video — dropdown from input directory
  • start_time FLOAT — seconds
  • target_fps INT
  • required_frames INT

Outputs

  • IMAGE batch — decoded frames
  • frame_count INT

Boyo Video Cutter

File: boyo_video_cutter.py | Category: Boyo/Video

Removes specific frames from an image sequence, typically the overlapping frames at chunk junctions in looped video generation. Pairs directly with Boyo Video Length Calculator.

Inputs

  • images IMAGE — full frame batch
  • trim_positions STRING — frame indices to remove, e.g. "94,95,96|188,189,190" (pipe separates junction groups)
  • debug_mode BOOLEAN

Outputs

  • trimmed_images IMAGE
  • cut_info STRING
  • frames_removed INT

Boyo Video Length Calculator

File: boyo_video_length_calculator.py | Category: Boyo/Video

Looks up a hardcoded table to return the total frame count, number of loop iterations, and junction trim positions needed to produce a target video duration. Works in 5-second increments up to 60 seconds.

Inputs

  • target_seconds INT — 5 to 60, step 5

Outputs

  • total_frames INT
  • loops_needed INT
  • trim_positions STRING — feed directly into Boyo Video Cutter
  • info STRING

Boyo Video Paired Saver

File: boyo_video_paired_saver.py | Category: Boyo/Video

Renders an image batch to a video file via FFmpeg and saves a matching .txt prompt file alongside it. Supports optional audio muxing. Useful for building video-text pair datasets.

Inputs

  • images IMAGE — frame batch
  • enhanced_prompt STRING
  • folder_name / filename_prefix STRING
  • fps FLOAT
  • codec — libx264, libx265, or av1
  • quality — high, medium, or low
  • audio AUDIO (optional)

Outputs: none (output node)

Boyo Load Video Directory

File: BoyoLoadVideoDirectory.py | Category: Boyo/Video

Loads every video file in a specified directory as an image batch. Supports common formats (mp4, mov, avi, mkv, webm).

Inputs

  • directory STRING
  • image_load_cap INT — max frames to load
  • skip_first_images INT

Outputs

  • image IMAGE
  • mask MASK
  • int — frame count

Boyo Frame Counter

File: boyo_frame_counter.py | Category: Boyo/Video

Computes frames_processed = (counter × chunk_size) + offset. Feed the loop counter and your chunk size to get the correct frames_processed value for nodes like VHS Video Combine in looped video workflows.

Inputs

  • counter INT — current loop iteration
  • chunk_size INT — frames per iteration (e.g. 89)
  • offset INT (optional, default 0)

Outputs

  • frames_processed INT

Boyo Overlap Switch

File: boyo_overlap_switch.py | Category: Boyo/Video

Outputs a different overlap value depending on whether this is the first loop iteration or a subsequent one. Prevents the first generated chunk from overlapping nothing.

Inputs

  • counter INT
  • first_overlap INT — typically 0
  • subsequent_overlap INT — your optimal blend value (default 13)

Outputs

  • overlap INT

Boyo Watermarks

File: BoyoWatermarks.py | Category: BoyoNodes/video

Stamps a watermark image onto every frame of a video batch. The watermark is resized to specified dimensions, composited at a chosen corner with adjustable opacity.

Inputs

  • frames IMAGE — video frame batch
  • watermark IMAGE — single watermark frame
  • wm_width / wm_height INT
  • corner — bottom-right, bottom-left, top-right, top-left
  • opacity FLOAT (0–1, default 0.85)
  • padding INT — pixel gap from edge

Outputs

  • frames IMAGE

Looping & Flow Control

Boyo While Loop Start / End

File: BoyoBastardLoops.py | Category: Boyo/Loops

While loop nodes with custom execution management. The Start node initialises loop state; the End node evaluates the continue condition and re-queues the Start node if true. Handles arbitrary typed pass-through values.

Notes: These patch ComfyUI's execution engine to support genuine cycles. Use with care in complex graphs.

Boyo For Loop Start / End

File: boyo_for_loops_exact.py | Category: Boyo/Loops

For loop nodes that replicate EasyUse's loop logic without requiring EasyUse as a dependency. Includes an internal tagged cache to persist values across iterations.

Inputs (Start)

  • total INT — number of iterations
  • initial_value0–4 — any type, pass-through values

Outputs (End)

  • flow — loop control signal
  • value0–4 — values from last iteration

Boyo Loop Reset

File: boyo_loop_reset.py | Category: Boyo/Loops

Resets named loop counters to zero when a trigger signal arrives. Use at the end of a workflow to restart a bastard loop from the beginning.

Inputs

  • trigger ANY — completion signal (e.g. from a save node)
  • reset_mode — immediate or delayed
  • loop_id_1–3 STRING (optional) — named loop IDs to reset

Outputs

  • trigger ANY — pass-through

Boyo Loop Counter / Boyo Math Int / Boyo Compare

File: boyo_for_loops_exact.py | Category: Boyo/Loops

Utility nodes for loop arithmetic. Loop Counter increments on each pass. Math Int performs addition, subtraction, multiplication, and division on integers. Compare outputs a boolean from two integer inputs.

Boyo Prompt Loop

File: BoyoPromptLoop.py | Category: Boyo/Loops

Iterates through prompts stored in .txt files in a prompts/ subfolder. Three modes: sequential (one per loop pass), random (seeded), or single (fixed index).

Inputs

  • text_file — dropdown of .txt files in prompts/ directory
  • mode — sequential, random, or single
  • start_seed INT
  • index INT — used in single mode

Outputs

  • prompt STRING
  • index INT — current position

Boyo Loop Collector / Boyo Loop Image Saver

File: BoyoLoopCollector.py | Category: Boyo/Loops

Loop Collector accumulates images across loop iterations into a growing batch. Loop Image Saver writes each collected image to disk with sequential filenames as they arrive.

Latent & Conditioning

Boyo Latent Cache Updater

File: boyo_latent_cache_updater.py | Category: Boyo/Latent

Writes a latent tensor into the Boyo loop cache keyed by a string ID, then passes the tensor through unchanged. Avoids graph cycles by separating the write from the next iteration's read (handled by Boyo Latent Switch).

Inputs

  • latent LATENT
  • cache_key STRING — must match the key used in Latent Switch
  • counter INT (optional)

Outputs

  • latent LATENT — unchanged pass-through

Boyo Latent Switch

File: boyo_latent_switch.py | Category: Boyo/Latent

On iteration 0 outputs start_latent; on all subsequent iterations reads the cached latent written by Boyo Latent Cache Updater. This is the clean way to feed the previous frame's latent back into a sampler without a cycle.

Inputs

  • counter INT
  • start_latent LATENT — used on first pass
  • cache_key STRING — must match Cache Updater key
  • next_latent LATENT (optional)

Outputs

  • latent LATENT

Boyo Latent Passthrough / Boyo Execution Barrier

File: boyo_latent_passthrough.py | Category: Boyo/Latent

Passthrough passes a latent through with no modification. Execution Barrier adds an explicit dependency edge — use it to enforce a specific ordering between otherwise unconnected graph branches.

Boyo Painter SVI

File: boyo_painter_svi.py | Category: conditioning/video_models

Merges PainterI2V motion amplitude conditioning with WanImageToVideoSVIPro context preservation. Intended for samplers 2 and onwards in infinite-length video generation, after the first PainterI2V pass has established motion.

Inputs

  • positive / negative CONDITIONING
  • length INT — frame count for this chunk
  • anchor_samples LATENT — latent from the first PainterI2V sampler
  • motion_amplitude FLOAT (1.0–2.0)
  • motion_latent_count INT
  • prev_samples LATENT (optional) — previous chunk for context blending

Outputs

  • positive / negative CONDITIONING
  • latent LATENT

Boyo VACE Injector

File: BoyoControl.py | Category: Boyo/Control

Injects VACE control data directly into model attributes, bypassing conditioning entirely. Supports standard ComfyUI MODEL or WanVideoWrapper's WANVIDEOMODEL.

Inputs

  • control_image IMAGE
  • vace_strength FLOAT (0–2)
  • vace_start_percent / vace_end_percent FLOAT
  • num_frames INT
  • model MODEL (optional)
  • wanvideomodel WANVIDEOMODEL (optional)
  • vae VAE (optional)

Outputs

  • model MODEL

Boyo VACE Viewer

File: BoyoControl.py | Category: Boyo/Control

Reads VACE control attributes from a model and displays them as a formatted string for debugging injection results.

Inputs

  • model MODEL

Outputs

  • info STRING

Utilities

Boyo Resolution Calc

File: BoyoResolutionCalc.py | Category: Boyonodes

Takes a base width and a named aspect ratio and returns width and height both rounded to the nearest multiple of 8.

Inputs

  • width INT
  • aspect_ratio — 1:1 Square, 4:3, 3:2, 16:9 HD, 21:9, 2:3, 3:4, 9:16 Tiktok-Shorts

Outputs

  • width INT
  • height INT

Boyo Apply LUT

File: boyo_lut.py | Category: BoyoNodes/Colour

Applies a colour-grading LUT to an image. Scans the luts/ subfolder at startup for .cube, .3dl, and .spi3d files.

Inputs

  • image IMAGE
  • lut_file — dropdown from luts/ folder
  • strength FLOAT — blend between original and graded (0–1)

Outputs

  • image IMAGE

Dependency: pip install colour-science

Setup: Drop your LUT files into ComfyUI/custom_nodes/Boyonodes/luts/.

Boyo Mask To Image

File: boyo_mask_normalise.py | Category: BoyoNodes/Masks

Converts a mask tensor of any standard shape to a 3-channel IMAGE tensor. Written specifically to work around DiffuEraser's broken len(video_mask) > 3 check, which incorrectly rejects any video longer than 3 frames.

Inputs

  • mask MASK — accepts [h,w], [b,h,w], [b,h,w,1], or [b,1,h,w]

Outputs

  • image IMAGE — shape [b,h,w,3]

Boyo Asset Grabber Simple

File: boyo_asset_grabber_simple.py | Category: Boyo/Utility

Reads a JSON manifest and automatically installs custom nodes (via git clone), Python packages (via pip), and downloads model files. Detects ComfyUI's base path automatically.

Inputs

  • manifest_json STRING — JSON config defining assets to install
  • execute BOOLEAN — dry-run when false

Outputs

  • log STRING — installation report

Boyo Asset Grabber Advanced

File: boyo_asset_grabber_advanced.py | Category: Boyo/Utility

Same as Simple but with explicit path override inputs for custom nodes directory, models directory, etc. Use when ComfyUI is installed in a non-standard location.

Boyo Prompt Relay Encode

File: boyo_prompt_relay_nodes.py | Category: (conditioning)

Port of kijai's ComfyUI-PromptRelay into Boyonodes. Encodes a set of prompts with temporally-varying Gaussian segment masks so different prompts influence different time steps of a video generation.

Inputs

  • clip CLIP
  • prompts — list of text segments
  • segment_lengths — duration of each segment as a proportion of total steps

Outputs

  • conditioning CONDITIONING

Attribution: Original logic by Jukka Seppänen (kijai), ported with attribution.

Boyo Prompt Relay Encode Timeline

File: boyo_prompt_relay_nodes.py | Category: (conditioning)

Timeline-based variant of Prompt Relay. Accepts explicit step-range definitions per prompt segment instead of proportional lengths, for finer control.

Boyo Prompt Relay LoRA Gate

File: boyo_prompt_relay_nodes.py | Category: (conditioning)

Applies a LoRA selectively to a single temporal segment using the same Gaussian gating mechanism as Prompt Relay. Enables LoRA influence to be confined to a specific portion of the video.

Inputs

  • model MODEL
  • lora_name STRING
  • strength FLOAT
  • segment_start / segment_end FLOAT — 0–1 proportional range

Outputs

  • model MODEL

For installation instructions and one-line node summaries see README.md.