Project Category: ComfyUI

A collection of practical ComfyUI utility nodes designed to reduce workflow complexity. Each node solves a specific recurring problem — avoiding VRAM crashes, picking prompts deterministically, filtering incompatible LoRAs, or labelling model pipelines. No external dependencies beyond the standard ComfyUI environment.

Nodes

Node	Category	Purpose
Mega Lora Loader	Oli/loaders	Stackable multi-LoRA loader with automatic compatibility filtering
Mega String List	Oli/prompt	Collects strings and lists from multiple sources into a unified list
Prompt Line Pick	Oli/prompt	Seed-driven line picker — uniform distribution, fully independent instances
Video Frame Limit	Oli/utils	Caps video duration to avoid VRAM out-of-memory crashes
Model Info	Oli/utils	Returns architecture details of any connected model
Node Label	Oli/utils	Reads and passes through the upstream node’s title

Installation

Via ComfyUI Manager (recommended): search for Oli Prompt Tools and click Install.

Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/magicoli/oli-prompt-tools

Restart ComfyUI. No additional dependencies required.

---

Mega Lora Loader (Oli)

Stackable multi-LoRA loader inspired by rgthree’s Power Lora Loader. Key additions: stackable, and each LoRA’s safetensors header is checked against the connected model’s key map before loading — incompatible LoRAs are silently skipped rather than producing errors or corrupted output.

• Add as many LoRA rows as needed with the ➕ Add LoRA button.
• Each row has an on/off toggle, a LoRA selector, and a strength slider.
• Incompatible LoRAs are highlighted in the UI (orange = incompatible, grey = disabled).
• Connect the enable input to a boolean condition to bypass all LoRAs at once — useful with switch or router nodes.
• Two usage modes: connect model/clip to apply LoRAs immediately, or leave them unconnected to only build a LORA_STACK without applying it. In stack-only mode the node acts as a LoRA definition block — chain several together, then feed the final stack to a downstream loader that has model/clip connected. This lets you define LoRAs once and apply them to multiple models at different points in the workflow.

Mega Lora Loader example workflow

Inputs

Input	Type	Default	Description
lora_stack	LORA_STACK	—	Incoming stack from another loader (optional)
model	MODEL	—	Model to apply LoRAs to (optional)
clip	CLIP	—	CLIP encoder to apply LoRAs to (optional)
(lora rows)	—	—	Added dynamically via the ➕ button
enable	BOOLEAN	true	When false, passes model/clip/lora_stack through unchanged

Outputs

Output	Type	Description
lora_stack	LORA_STACK	Full stack including upstream entries
MODEL	MODEL	Model with all compatible LoRAs applied
CLIP	CLIP	CLIP with all compatible LoRAs applied

---

Mega String List (Oli)

Collects strings and lists from multiple sources into a single unified list. Each row can hold either typed text or a connected node output (STRING, LIST, or any type) — no separate input slots needed for typed vs. connected content. Rows can be individually toggled on/off, reordered by drag, or deleted. Lists from connected nodes are automatically expanded before addition.

• Mixed sources — combine typed strings, node outputs, and upstream prompt lists in any order in a single node.
• Drag to reorder — grab the ≡ handle to change the order of rows; the final list follows the visual order.
• Per-row toggle — disable individual rows without deleting them using the pill switch.
• Chainable — wire optional_prompt_list from a Prompt Line Pick or another Mega String List; the incoming list is prepended to the result.
• Pass-through mode — when enable is set to False (labelled pass-through), the node returns only optional_prompt_list unchanged and ignores all rows. Useful for temporarily disabling additions or for conditional branching.
• Delimiter — used both to split typed multi-value text into separate items, and to join all items into the string output. Supports escape sequences (n, t, …).
• Items equal to "none" (case-insensitive) are automatically removed — SDXL Prompt Styler’s empty/ignore sentinel.

Mega String List example workflow

Inputs

Input	Type	Default	Description
optional_prompt_list	LIST	—	Accumulated list from an upstream node — prepended to the result (optional)
enable	BOOLEAN	true	When false (pass-through): return only optional_prompt_list, ignore all rows
delimiter	STRING	,	Splits typed text into items; also used to join the string output
string 1-n	*	—	Dynamic rows — type text directly or connect any node output; lists are expanded

Outputs

Output	Type	Description
prompt_list	LIST	Combined list — wire to another Mega String List or Prompt Line Pick
prompt_strings	STRING (list)	Same items as a STRING output list — compatible with easy promptList
num_strings	INT	Number of items in the combined list
string	STRING	All items joined by the delimiter

---

Prompt Line Pick (Oli)

Reimplementation of the easy promptLine concept. Replaces start_index with a seed. The picked index is derived via sha256(seed:node_id) % len(lines), giving a uniform distribution and full independence between instances: two pickers with the same seed pick at uncorrelated positions even when their lists have the same length or lengths that are multiples of each other. Output format is identical to easy promptLine — both STRING and COMBO return the list starting at the picked line.

• Uniform distribution — every line has equal probability regardless of list length, with no correlation between lists of similar sizes.
• Full independence — multiple instances in the same workflow each pick at independent positions, even with the same seed, because the node ID is part of the hash.
• COMBO output is compatible with any COMBO-typed input (e.g. SDXL Prompt Styler artist/style fields): the picked line is always the first element.
• Stackable — connect optional_prompt_list from a previous picker; this node appends its pick and passes the extended list through prompt_list. Chain as many pickers as needed, then feed into easy promptList or any string-join node.

Prompt Line Pick example workflow

Inputs

Input	Type	Default	Description
prompt	STRING	—	One item per line
seed	INT	0	Workflow seed — share with KSampler for reproducible pairs
remove_empty_lines	BOOLEAN	true	Strip blank lines before picking
uncorrelate	BOOLEAN	true	When on: index = sha256(seed:node_id) % len — independent between instances. When off: index = seed % len — same as easy promptLine’s start_index
optional_prompt_list	LIST	—	Accumulated list from an upstream picker (optional) — same type as easy promptList

Outputs

Output	Type	Description
STRING	STRING	The single picked line (scalar)
COMBO	COMBO (list)	The single picked line, compatible with COMBO-typed inputs (SDXL Prompt Styler etc.)
prompt_list	LIST	Incoming list with this pick appended — wire to the next picker or to easy promptList
prompt_strings	STRING (list)	All strings in the accumulated list — same format as easy promptList’s prompt_strings
seed	INT	Pass-through — wire to the next picker without routing back

---

Video Frame Limit (Oli)

Caps video generation duration to avoid VRAM out-of-memory crashes. The frame budget is derived from transformer peak memory first principles rather than empirical constants:

bytes_per_latent_frame = TENSOR_COPIES × (width÷8) × (height÷8) × hidden_dim × 2
max_frames             = total_vram × safety_margin ÷ bytes_per_latent_frame

Where TENSOR_COPIES = 5 (Q, K, V, attention output, residual activations) and hidden_dim is auto-detected from the connected model. Uses total VRAM rather than free VRAM — ComfyUI offloads weights layer-by-layer, so peak activation memory scales with total VRAM, not the remainder after model loading.

The node displays detected VRAM, model name, hidden dim, requested and capped frames directly on the canvas after each execution — making it usable as a standalone config panel for the whole generation.

Video Frame Limit example workflow

Inputs

Input	Type	Default	Description
width	INT	832	Generation width in pixels
height	INT	480	Generation height in pixels
fps	FLOAT	16	Frames per second
duration	FLOAT	10	Requested duration in seconds
safety_margin	FLOAT	0.95	Fraction of total VRAM to budget (0.95 = 5% headroom)
model	MODEL	—	Optional — enables hidden_dim auto-detection

Outputs

Output	Type	Description
width	INT	Pass-through
height	INT	Pass-through
frames	INT	Capped frame count
fps	FLOAT	Pass-through
duration	FLOAT	Actual duration after capping

---

Model Info (Oli)

Returns architecture details of any connected model (MODEL, CLIP, VAE, or any other type). Also reads the upstream node’s title. Useful for debugging model pipelines, inspecting what arrived on a bus, or driving conditional logic based on model class.

Model Info example workflow

Inputs

Input	Type	Description
model	*	Any model type (optional)

Outputs

Output	Type	Description
class_name	STRING	Python class name of the model (e.g. WAN21, Flux)
dim	INT	Hidden dimension auto-detected from the model
label	STRING	Title of the upstream node that produced the model

---

Node Label (Oli)

Passes any value through and outputs the title of the upstream node at a configurable traversal depth. Useful for labelling outputs, routing between branches, or building self-documenting workflows where node titles carry semantic meaning.

Node Label example workflow

Inputs

Input	Type	Default	Description
depth	INT	1	Hops to travel upstream (1 = direct parent, 2 = grandparent…)
node	*	—	Any value — passed through unchanged (optional)

Outputs

Output	Type	Description
node	*	Pass-through of the input value
label	STRING	Title of the node depth hops upstream

---

License

GNU Affero General Public License v3.0