ComfyUI_Lam

AppParams and the WeChat app platform hiding inside this pack

30 resolutions, zero math

Pad your audio with silence, before or after, in seconds

Get an audio file's real specs

AutioPath and the typo nobody fixed

The condition-check end of the do-while loop that needs a patched ComfyUI

The do-while loop opener that hands you a token, a counter and a seed

EasyPromptSelecto and the lam tag library

The GAN-based face fusion node — model from a Baidu Pan link, quality worth the hoops

The counted-loop tail that re-queues your graph until it's done

The in-process loop tail that collects every iteration's result into one batch

The patched, in-process loop head with a data backhaul

Total, step, and a seed that changes every pass

Read a QR code back to text without leaving ComfyUI

IfInnerExecute

Turn a still into a KLing AI clip, with real camera moves

Burn a mask into an image's alpha channel so composites don't fight you

A solid-color image generator — the humble node every canvas starts with

Need more frames? ImageClone duplicates your batch, on purpose

Crop every face in a photo into clean squares, masks and boxes included

ImageLama's prompt-free inpainting

Turn a color channel into a mask — batch-friendly, one drop-down

Assemble your clips into a 剪映/CapCut-ready video — the mid-graph variant

The 剪映/CapCut compose node that finishes your workflow properly

Turn a topic into a video prompt, courtesy of DeepSeek

A node with no inputs that outputs 'anything'

The model-list picker — enumerate any ComfyUI folder and pull one name out

Print anything to the console — and keep the value flowing

Print mid-graph without claiming the workflow is done

The face-analysis loader that powers the pack's crop nodes — insightface or auraface

Read a PNG's embedded workflow from inside your graph

HeyGem lip-sync from inside ComfyUI (without running the model yourself)

Check a HeyGem lip-sync job without blocking the queue

IndexTTS2 advanced knobs, bundled into one node you can reuse

IndexTTS2 voice cloning with timbre and emotion from one reference clip

IndexTTS2 with the voice from one clip and the emotion from another

IndexTTS2 voice cloning with eight dials of emotion instead of a reference clip

IndexTTS2 voice cloning with emotion described in plain words

Free your VRAM when you're done with IndexTTS2

Paste an image into ComfyUI as base64 text

Load an image from an absolute path or a URL — not just the input folder

Pull video frames into the graph, with optional audio extraction

The node that wires MiniMax H3 reference-to-video together

The LAM pose editor

Turn an equirectangular panorama into a spherical (fisheye) view

Filter a folder's files by keyword and format, with AND/OR/negation

One dropdown to drive every sampler in your workflow

Export your generated audio as an mp3 with a quality knob

Save an image and keep the graph flowing (no metadata faff)

Save a VIDEO tensor to disk, with format and codec control

Save your video and get the path back — without ending the workflow

Pick a scheduler from a dropdown and wire it anywhere

Stitch a spherical view back into a full equirectangular panorama

A lazy case/switch for arbitrary ComfyUI values

Toggle a segment of your pipeline on or off from one control point

A master on/off for any branch of your graph

Save and preview video straight from your video workflow

The near-twin of LamViewVideo, with a bonus path output

Turn a folder of images into a list of paths

Why you can't just use LoadImage here (and what LoadImagePath does instead)

Paste a patch back onto the canvas — LoadReplaceImage's mask blend

One big text block in, an array out

Stack several ControlNets with per-unit strength

Draw boxes on the node, place subjects with GLIGEN

MultiIntFormula's expression evaluator

Mask-limited IP-Adapter, several references at once

Concatenate a list of audio files into one MP3

Stitch your clips into one file without leaving the graph

A Python eval box that sits right in your graph

Gluing text together in the lam pack

Batch-encode a list of prompts into one conditioning

Batch-encode a list of prompts with BNK-style weighting

Pick one of several stacked text inputs by index

Different prompts for different boxes on the canvas

Ground each prompt to a box with a GLIGEN model

Prompt-by-mask, the precise way to split a scene

An OpenAI-compatible chat node that defaults to Aliyun

The condition that keeps re-queuing your graph

The top of a do-while loop that ComfyUI never wanted

PreviewImageLam

Pipe Chinese prompts through Baidu Translate, mid-graph

Make the pattern, let ControlNet make the art

Qwen-Image editing in ComfyUI — but it's an API call, not a local model

Alibaba's Qwen-Image, but through the cloud, not your GPU

Save images mid-loop and get the file paths back

The SaveImage that bakes your own metadata into the PNG

The return trip for ComfyUI's DIY load-balancer

The front half of ComfyUI's DIY load-balancer

SDXL style presets, A1111-style, with a clickable tag UI

Free Microsoft TTS voices, no API key, straight to MP3

Pull one string out of a list, by index

Mux audio onto a video with ffmpeg, without leaving ComfyUI

Pry the audio track off a video, no re-encode

Frame-by-frame face swap, straight onto a video

How long is that video, actually? Ask VideoInfo

A file picker for videos, so you stop hand-typing paths

Make the workflow pause until you pick a winner

Translate prompts to 11 languages locally, no API key