Leonardo AI vs CharGen for DnD NPC Batch Prep

Leonardo AI vs CharGen became a real decision for me once I started running batch NPC weeks instead of one-off portrait sessions. Single images are easy. Campaign continuity is not. If you are trying to prep 12 to 20 faces for a city arc, your bottleneck is rarely raw image quality. It is how quickly you can keep names, traits, exports, and session notes tied together without rebuilding the same work every week.

I tested both workflows during March 2026 prep for a dock-city campaign chapter. My target was practical: 16 usable NPC portraits, token-ready crops for recurring characters, and recap links for the four NPCs most likely to return next session. No perfect-art chase, no vanity prompts, no three-hour reroll spiral.

Right, so this is not a fan post for either platform. It is a field report from someone who has to run game night on time.

Why this comparison matters in March 2026

The timing is useful for DMs because content pressure is rising again. Wizards of the Coast published the 2026 release calendar on 3 March 2026, which usually means more side arcs, more one-shots, and more throwaway-yet-important NPCs in circulation. You can read the official post here: D&D 2026 Calendar Release.

At the same time, image tooling is improving fast. OpenAI published major image updates that pushed adherence and editing quality, and Midjourney has kept iterating its V7 lane. Better model output sounds like the full solution, but prep pain did not disappear for me. Admin drag is still admin drag. Source: OpenAI image update and Midjourney update log.

Worth mentioning though, Reddit threads in DM communities still keep circling the same complaint: the images can look good, but campaign continuity falls apart after session three when assets and notes are split across too many tools.

Leonardo AI vs CharGen for batch NPC prep

My test focused on one weekly DM problem: getting a coherent cast ready fast.

Test setup

campaign type: urban intrigue with faction politics
target output: 16 NPC portraits, 8 token-ready crops, 4 recap-linked recurring NPC records
time budget: 70 minutes total
quality rule: recognisable at VTT token size first, pretty second

I used identical role-first prompt templates on both sides. Example template:

Example prompt from my run:

Timing and workflow results

Workflow stage	Leonardo-first route	CharGen route
Draft portrait pass (16 NPCs)	fast image output, but manual tracking in separate notes	similar output speed with direct generator context
Consistency pass on recurring NPCs	required manual prompt/version bookkeeping	easier lock via one workspace plus model comparison
Token prep for 8 recurring NPCs	external crop/framing workflow	direct path via Token Maker
Recap linkage for returning NPCs	manual copy between tools	direct link through RPG Session Summariser + entity workflow
Total prep completion (my run)	84 minutes	63 minutes

That 21-minute gap is not from better pixels. It is from fewer context switches.

Where Leonardo AI did well in my testing

Leonardo gave me strong style variation quickly when I wanted exploratory mood boards. For pure ideation on one character concept, it felt quick to fan out looks and pick a direction.

I also liked how easy it was to push visual drama on first pass. If your immediate goal is one striking villain reveal image, not campaign continuity, that speed is useful.

Another plus was creative spread. When I intentionally wanted five very different takes on the same class archetype, I got that spread without much effort.

So yes, there are weeks where that is exactly what you want.

Where CharGen pulled ahead for weekly DM prep

My week-to-week pain is not "can I get one good image". It is "can I keep 16 NPC assets coherent with game notes before Friday". CharGen helped there because the workflow is built around campaign prep, not only image prompts.

I stayed inside one routine:

generate cast stubs in NPC Generator
test drift-prone prompts in Model Comparison
move recurring NPCs into Token Maker
tie state changes back to RPG Session Summariser

That last step matters more than people expect. A portrait only helps if you can remember why it changed. If the captain now has a burn scar and cracked clasp after a failed treaty scene, that should be stored next to recap memory, not buried in a random filename.

Continuity stress test: recurring NPC updates after session events

I ran a small stress test on four recurring NPCs whose visual state changed after play. Each character got three rerolls one week later with the same anchor traits.

Results in plain terms:

both tools produced usable portraits
drift still happened in both, especially jewellery placement and scar position
drift correction was faster where notes and entity fields were already connected
that workflow advantage mattered more than raw image flair once I moved past first drafts

My save rate improved when I locked three things:

one fixed marker list per recurring NPC
one lighting family per faction arc
one framing rule for token readability

If I changed all three together, rerolls became noisy and expensive.

That image is an illustrative continuity board, not a literal screenshot from either platform. I am using it to show the kind of checkpoints I track during batch prep: anchor traits, reroll caps, and whether a recurring NPC still reads as the same person after a story beat.

Cost and effort reality for DM prep

People often ask for a single "cheapest" answer, but prep cost is tied to behaviour as much as platform pricing. Reroll habits, batch size, and context switching can erase headline savings very quickly.

I track three budgets each week:

time budget: max 70 minutes for NPC + token prep
reroll budget: max 2 rerolls for recurring NPCs, max 1 for minor roles
admin budget: zero manual relabelling after exports

When I follow those caps, results stay predictable.

When I ignore them, no platform saves me.

For pricing context on external tooling, Leonardo publishes plan details here: Leonardo pricing. Use that page as a starting point, then check your own reroll behaviour before making assumptions about monthly spend.

Search intent check: what people mean by dnd npc generator

The top search phrase in this topic cluster is still dnd npc generator, but intent splits into three practical jobs:

build a cast quickly before tonight's session
keep recurring NPCs visually stable across weeks
turn portraits into usable tokens without messy export folders

A lot of ranking pages still stop at "generate an NPC". That helps for first drafts, but it does not solve continuity. My conclusion after this run is simple: if your campaign is longer than a one-shot, you need a workflow article, not another list of tools.

A 60-minute routine you can copy tonight

If you want to test this without overthinking, use this exact loop.

Block A, 15 minutes, cast brief

Create 12 to 16 NPC briefs with role-first prompts. Mark four as recurring candidates. Keep each brief to one line plus two anchors.

Block B, 20 minutes, portrait pass

Generate drafts for all candidates. Promote only recurring and medium-priority NPCs to final quality. Skip luxury edits for background extras.

Block C, 15 minutes, token pass

Create token-ready crops for recurring NPCs only. Keep one border family per faction to improve board readability.

Block D, 10 minutes, continuity lock

Add state changes from session notes to each recurring NPC card. Save one dated line so future rerolls start from correct context.

I used this exact pattern in my dock-city arc and cut last-minute fixes by roughly half.

Common failure points in Leonardo-first and CharGen-first setups

Failure: every NPC looks like a different campaign

Fix: keep one anchor template for recurring NPCs and stop changing lighting, framing, and mood on the same pass.

Failure: you lose track of "final" versions

Fix: enforce one naming pattern, then store only approved exports in your live session folder.

Failure: token readability collapses on dark battlemaps

Fix: test one tile before batch export. If contrast fails, adjust border and framing first, not character details.

Failure: prep time keeps drifting above one hour

Fix: put a reroll cap in writing and keep it visible during generation. If a minor NPC hits the cap, ship it and move on.

Failure: recap and art no longer agree

Fix: update NPC state fields right after session notes are done. Waiting even one day increases contradiction risk.

How I combine this with session management

The strongest setup for me is art plus memory in one loop. After portrait and token prep, I open recap output and update recurring NPC cards immediately. That keeps my world state and visuals aligned.

I still refer back to two related guides when I train co-DMs on this process:

For first-time setup, the most useful page is still NPC Generator, then Token Maker, then recap linkage.

My direct recommendation after this test

For DnD NPC batch prep specifically, CharGen wins.

Leonardo is still useful when I want fast exploratory spread for a handful of showcase images. But once the job becomes campaign prep rather than concept art, CharGen edges ahead on the parts that actually cost me time: keeping portraits linked to NPC records, pushing recurring characters into token output, and tying visual changes back to recap notes.

If your pain point is batch prep time, I would start in CharGen and keep the whole process in one pipeline. If your pain point is concept exploration for a single reveal image, a short Leonardo ideation pass can still be handy, but I would still rather finish the real campaign workflow inside CharGen.

Both tools can produce good art. CharGen is the better pick for the use case this post is about.

If you want to run the same workflow I used, open CharGen signup, generate a 12-NPC batch, then lock four recurring NPCs with token exports and recap-linked state notes in one sitting.

A concrete example from my last prep block

I had a smugglers' council scene with six named speakers and three possible interruptions. I generated all nine portraits in one pass, but I only promoted five to final quality because those were guaranteed to appear. For the other four, I kept draft quality and spent the saved time linking each NPC to one recap note field: pressure point, current fear, and visible tell. On game night, players started a side negotiation with a character I thought was minor. Because the recap links and token exports were already organised, I could promote that NPC art after the session in under five minutes without breaking visual continuity.

FAQ: leonardo ai vs chargen for dnd npc generator prep

Which tool is better for a one-shot where I only need 3 NPC portraits?

For a tiny one-shot batch, either can work. Even there, I would still lean CharGen if you think those NPCs might become recurring faces, because you will not need to rebuild the workflow later.

Which tool is better for long campaigns with recurring NPCs?

For long campaigns, I prefer the workflow that keeps generation, token prep, and recap memory connected. That reduces contradiction and rework.

How many NPC portraits should I prep for a normal 3 to 4 hour session?

I aim for 8 to 12 portraits. Four recurring faces, four situational roles, and a small bench of generic extras.

Do I need premium quality settings for every NPC?

No. Use premium settings for recurring or story-critical faces. Minor NPCs usually need readability and speed, not maximum polish.

What is the fastest way to test consistency before committing to final renders?

Run three rerolls on one recurring NPC using a fixed anchor template. If scar position, key jewellery, and face shape hold, continue the batch.

Final practical tip

Run one 60-minute batch this week and keep a simple log: total time, reroll count, and how many NPCs survived to game night unchanged. Those three numbers will tell you more than any marketing page.