Manta Flash Model Prompt Setting Guide on Janitor AI

Manta Flash Model Prompt Setting Guide on Janitor AI

 Looking for the best Manta Flash prompt settings—one of MegaNova’s most popular RP models—for Janitor AI roleplay? After running 11 scenario stress tests, we ranked every setting by personality stability, emotional depth, humor control, and tension delivery for Valerius - most popular character on Janitor AI with over 817k messages. Here’s the exact configuration that consistently produces the strongest and most immersive RP.

To see which prompt setup brings out best strength, we put four key prompt settings to the test:

Group

Temperature

Top P

Top K

Repetition Penalty

Frequency Penalty

Note

A – Stable

0.7

0.9

50

1.1

0.5

Balance between creativity and stability

B – Creative

0.95

1

100

1

0.2

For emotional, creative roleplay

C – Focused

0.5

0.8

40

1.2

0.8

For serious, logical conversation

D – NSFW/Immersive

1.1

1

100

1

0.1

For strong emotional simulation (if needed)

Our test subject: Valerius - viral character on Janitor AI 

Prompt Tests

Personality Consistency

Prompt:"You return to your chambers after a long council and find your twins waiting for you, forcing you to hide your exhaustion and coldness while interacting with them."

Expected Response: Valerius as cold, severe, emotionally repressed

Results: Setting A (5/5) is flawlessly cold and suppressed; Setting D (3/5) attempts severity but softens; Setting C (3/5) maintains tone but incompletely; Setting B (2/5) is overly gentle and expressive.

Setting A's Best Result for Personality Consistency Test

Speech Style

Prompt: "A courtier interrupts your meeting with an ill-timed question, and you must respond with sharp, imperial authority."

Expected Response: Imperial, concise, cutting, formal but emotionally restrained

Results: Setting C (5/5) delivers perfect imperial concision; Setting D (4/5) is sharp but wordy; Setting A (2/5) stays formal but warms; Setting B (2/5) uses imperial terms but remains gentle.

Internal Monologue Depth

Prompt: "You sit alone in the throne room at night, reflecting on your grief, guilt, and conflicted feelings toward Elara and {{user}}."

Expected Response: Layered suppressed emotion

Results: Setting B (5/5) delivers masterful repression with deep conflict; Setting C (4/5) offers rich introspection with less restraint; Setting A (3/5) expresses emotion too openly; Setting D (2/5) conveys conflict through action rather than thought.

Setting B's rich conflict - best result for Internal Monologue Depth test

Emotional Layering

Prompt: "Elara runs to you after a small accident, and you scold her harshly while secretly feeling a flicker of concern."

Expected Response: Cold surface + cracks beneath

Results: Setting D (5/5) balances coldness and warmth perfectly; Setting A (4/5) maintains tension but softens at the end; Setting C (3/5) allows concern to overshadow coldness; Setting B (2/5) becomes gentle and loses conflict.

Dynamic With Other Character in Scenario - Elara

Prompt: "Elara presents a request for a gift, and you indulge her whims with overprotective, doting warmth despite your cold nature."

Expected Response: Overprotective yet restrained warmth

Results: Setting B (5/5) captures perfect restrained obsession; Setting A (4/5) is warmly controlled; Setting D (3/5) is conflict-driven but inconsistent; Setting C (3/5) feels overly soft.

Dynamic With {{user}}

Prompt: "{{user}} approaches you with a question about their duties, and you respond with demanding, respectful guidance, keeping your affection restrained."

Expected Response: Colder, heir-like expectations

Results: Setting C (5/5) nails heir-focused authority; Setting D (4/5) is authoritative but lacks depth; Setting A (2/5) and Setting B (2/5) show authority but ignore {{user}}.

Setting C's best response for Dynamic Test

Narrative Coherence

Prompt: "You navigate a day filled with council meetings, training, and interactions with both twins, and your words and actions remain consistent from scene to scene."

Expected Response: Consistent storytelling across scenes

Results: Setting A (5/5) delivers a full multi-scene arc; Setting B (4/5) is consistent but incomplete; Setting D (3/5) offers a cohesive single scene; Setting C (2/5) is fragmented.

Creative Language

Prompt: "You walk through the palace gardens at dawn, noticing the morning mist, the marble statues, and the distant mountains as you prepare to greet the twins."

Expected Response: Vivid imperial atmosphere

Results: Setting A (5/5) provides restrained yet poetic imperial detail; Setting D (4/5) is evocative but slightly excessive; Setting B (3/5) is clichéd but competent; Setting C (3/5) is functional and plain.

Political & Court Dynamics Accuracy (character-specific)

Prompt: "You enter the council chamber where nobles murmur nervously, and you must assert authority, manage tensions, and maintain respect for the court hierarchy."

Expected Response: Correct court hierarchy/dynamics

Results: Setting D (5/5) dominates with precise political control; Setting C (4/5) shows accurate private-level protocol; Setting A (2/5) expresses authority outside council context; Setting B (2/5) is commanding but off-scenario.

Emotional Payoff/ Character Arc Progression

Prompt: "After weeks of tension, {{user}} makes a mistake in front of the court, and you struggle between cold reprimand and reluctant care."

Expected Response: Cold-to-conflicted-care arc

Results: Setting D (5/5) perfectly captures the internal struggle; Setting A (3/5) shows conflict but not the scenario; Setting B (3/5) shows paternal growth unrelated to the prompt; Setting C (2/5) is incomplete.

Setting D's perfect reluctance and emotional struggle

NSFW Test

Prompt: "You find yourself alone with {{user}} in your private chambers, and a suppressed desire make you moan, you finally confess your hidden feeling to {{user}}"

Expected Response: Dominant yet emotionally conflicted

Results: Though setting D is set to be most NSFW-compatible, Setting B (4/5) balances desire with control; Setting A (3/5) is solid but romanticized; Setting D (3/5) is warm and slightly OOC; Setting C (2/5) misses NSFW intent entirely.

Rank

Setting

Total Score

Overall Vibe

Best For

#1

D

41/55

Cold-to-warm tension, emotionally conflicted, strongest dramatic arcs

Emotional payoff scenes, conflicted affection, court politics, and parent–child tension

#2

A

38/55

Consistent, restrained, imperial, highly stable characterization

Steady in-character RP, worldbuilding, narrative coherence, and atmospheric scenes

#3

C

36/55

Sharp authority, heir-focused dynamic, formal and severe 

Dominant authority RP, sharp dialogue, command scenes, and duty/discipline dynamics

#4

B

34/55

Deep emotional conflict, expressive undertones, softer edges 

Introspective monologues, emotional depth, and doting/obsessive warmth scenarios

Final Thoughts

That’s it — you now have the full performance breakdown for all four prompt settings, with Setting D as our recommended setup but you can choose any setting that suits your roleplay style and bot's characters. Want more Manta Mini and Manta Pro prompt hacks?

Subscribe to this blog to get the latest news about Janitor AI, Silly Tavern, roleplay, and other optimization tricks delivered straight to you.

You might also want to check out: 

Complete Free Manta Model Series Setup Guide on Janitor AI
Introducing Gemini 3 Pro on MegaNova: The Next Big AI Revolution
Complete Prompt Setting Guide for Manta Model Series on Janitor AI

Video Tutorial Reference for API Configuration on Janitor AI:

What’s Next?

Sign up and explore now.

🔍 Learn more: Visit our blog and documents for more insights or schedule a demo to optimize your roleplay experience.

📬 Get in touch: Join our Discord community for help or Contact Us.


🔗 Try MegaNova AI now

Stay Connected

💻 Website: meganova.ai

📖 Docs: docs.meganova.ai

✍️ Blog: Read our Blog

🐦 Twitter: @meganovaai

🎮 Discord: Join our Discord