Asset Selector v3.6 (Prod) - Akki
The production manager that parses your shot list CSV
- master_character_list
- master_prop_list
- master_set_dressing_list
- master_costume_list
- master_vfx_list
- master_sfx_list
- total_shots_count
- total_character_count
- selected_character_name
- selected_character_costumes
- debug_output
- master_main_sets_list
- total_main_sets_count
- selected_main_set_name
- set_hierarchy_json
The Asset Selector v3.6 is the one node in the AkkiNodes pack that doesn't call an LLM at all - and that's the point. It's pure deterministic Python that parses the production CSV and hands you master lists of every asset in the film: characters, props, set dressing, costumes, VFX, and SFX, plus counts and the ability to select a specific character or set. This is the "code is law" philosophy at its purest: this is bookkeeping, and bookkeeping should never be left to a stochastic model.
What it does
Feed it the canonical production CSV from Phase 2 and it returns a wall of outputs. The ones you'll actually reach for:
master_character_list,master_prop_list,master_set_dressing_list,master_costume_list,master_vfx_list,master_sfx_list- every asset, deduplicated across the whole film.total_shots_countandtotal_character_count- quick production stats.selected_character_nameandselected_character_costumes- the character chosen by thecharacter_selectorindex, and their costumes.selected_main_set_nameandset_hierarchy_json- the set chosen byset_selector, plus a JSON hierarchy of sets and their variations. This feeds the Set Lookdev node directly.master_main_sets_listandtotal_main_sets_count- the unique location sluglines treated as master sets.
The v3.6 design treats each unique location slugline as one master set and only extracts time-of-day as a variation - a deliberately simple, robust structure so downstream nodes don't have to untangle a mess of near-duplicate sets.
The inputs
Just three: csv_report (the STRING containing your CSV), character_selector (an integer, default 1), and set_selector (an integer, default 1). The selectors are 1-based indexes into the character and set lists. Bump them to move through the cast and locations - in a loop, this is how you'd iterate over every character for lookdev.
Install
Standard pack install, no extra dependencies beyond the pack itself:
cd ComfyUI/custom_nodes
git clone https://github.com/routhakash/AkkiNodes-LLM-Suite-for-ComfyUI
Or ComfyUI Manager → search "AkkiNodes LLM Suite" → install, run install.bat as admin on Windows, restart. No GGUF needed for this node since it never touches a model - though the pack as a whole does.
One thing worth knowing: the source uses hard-coded Python alias dictionaries to merge asset synonyms (e.g., "sturdy boots" and "hiking boots" count as one costume). The author's own bug log flags this as a candidate to move into a user-editable aliases.json, but for now, if your CSV uses wording the aliases don't cover, you'll see duplicates in the lists. Cosmetic, mostly - but it's the kind of thing that makes you check the console once before trusting the counts.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| csv_report | STRING | — | |
| character_selector | INT | 1 | — |
| set_selector | INT | 1 | — |
Outputs (15)
| Name | Type | Description |
|---|---|---|
| master_character_list | STRING | — |
| master_prop_list | STRING | — |
| master_set_dressing_list | STRING | — |
| master_costume_list | STRING | — |
| master_vfx_list | STRING | — |
| master_sfx_list | STRING | — |
| total_shots_count | INT | — |
| total_character_count | INT | — |
| selected_character_name | STRING | — |
| selected_character_costumes | STRING | — |
| debug_output | STRING | — |
| master_main_sets_list | STRING | — |
| total_main_sets_count | INT | — |
| selected_main_set_name | STRING | — |
| set_hierarchy_json | STRING | — |