The two axes that matter
Every roleplay chatbot sits somewhere on two axes:
Hosted vs self-hosted. Hosted products (Sloane, Character.AI, Muah, Kindroid) work out of the box, no setup. Self-hosted stacks (SillyTavern + local LLM or API) give you total control but require hours of configuration.
Text-only vs multimodal. Text-only products deliver just chat. Multimodal products add photos (and sometimes voice, video) that stay consistent with the character's look. Multimodal is what turns "chat with a character" into "roleplay with a character."
Most products cover one axis (text-only + hosted, like Character.AI) or the other (self-hosted + text-only, like SillyTavern default). Sloane sits at hosted + multimodal, with the added structure of an actual Roleplay skill that has photos at scheduled beats through the arc.
The 2026 ranking
#1 — Sloane (Free / $9.99 Plus / $19.99 Premium) Hosted companion product with a dedicated Roleplay skill. You set the scenario, the persona plays into it, custom photos land at beats through the arc, and memory commits when you close. Per-persona LoRA keeps her face consistent. Free tier is real: 50 messages/day, 5 free custom photos, no card. Sandra is the most-picked starting persona but any of the ~80-persona roster works. Roleplay skill unlocked on Plus.
#2 — SillyTavern + local LLM or API (Free software, hours of setup) Open-source front-end for self-hosted or API-connected LLMs. Maximum control — model choice, sampler tuning, character-card customization. Real setup cost (4-12 hours cold-start) and needs either a gaming GPU for local or ongoing API spend. Best ceiling if you're technical. Full SillyTavern vs Sloane comparison.
#3 — Character.AI (Free with caps, $9.99 c.ai+) Massive character catalog (millions of user-created cards). Best if you want to try a specific niche character concept someone else already wrote. Real friction: daily message cap on free (50-80/day), character drift over long conversations, SFW-strict policy since the January 2026 wrongful-death lawsuit settlements and under-18 ban.
#4 — Muah AI ($9.99-24.99/mo, coin economy) Roleplay plus voice plus photo focus. Real strength: voice and photo integration. Real weakness: coin economy on top of subscription — every custom photo or voice minute costs coins. Realistic all-in monthly spend climbs fast. Good if you're specifically after voice-heavy roleplay.
#5 — Kindroid ($9.99-19.99/mo) Reddit-favorite for character depth. Strong memory system, personality-first design. Not roleplay-specialized — it's a companion product where roleplay is a use case, not a dedicated skill. Solid choice if character depth matters more than photo integration.
#6 — Janitor AI (Free with API key) Unfiltered by default (bring your own LLM API key). Character library similar to Character.AI's but without the content restrictions. Real friction: no built-in LLM (you pay for API access yourself), limited memory, no photos out of the box.
What "roleplay" actually means across these products
The word "roleplay" covers different things in different products:
Freeform text roleplay (Character.AI, Janitor AI, SillyTavern) — you set a scenario ("you're my new personal trainer"), the character plays into it, the conversation runs until you close it. No structured arc, no photos.
Structured skill-based roleplay (Sloane's Roleplay skill) — same setup, but the skill layer schedules photo delivery at beats through the arc (typically 3 custom photos at turns 2, 5, and 8), tracks scene state, and commits a memory summary when you close. The character remembers the roleplay after it ends.
Persistent character-driven (Kindroid) — no explicit "start roleplay" mode, but you can direct any conversation into a scenario and the character follows. Strong memory means the character brings up past roleplay moments unprompted.
Multimodal roleplay with voice (Muah, some Character.AI premium tiers) — text roleplay with voice-note delivery mid-scene. Real UX moment when a scene beat gets delivered as audio.
Sloane's Roleplay skill sits at the "structured + multimodal + persistent" intersection — the closest thing to "the character actually plays into the scenario with visual delivery at beats."
Photos in roleplay — why it matters
Roleplay is a scene-based medium. If you're running "we're at the vineyard on the Sunday of our trip," the moment where she sends a photo of herself at the vineyard is what turns the scenario from text-mode to felt-mode. That's the beat where a lot of roleplay users report the shift from "I'm reading a story" to "this is happening."
But photo delivery has an identity-consistency problem. If she sends a photo Sunday and another Monday and they're two different women, the scene collapses. This is why per-persona LoRA (Sloane) or IP-Adapter conditioning (some SillyTavern setups) matters — you need the same face across every photo for the scene to hold.
Most roleplay-focused products don't solve this well. Character.AI's image generation (when available) is generic. Muah's photos are consistent-ish but coin-metered. SillyTavern with a custom-trained LoRA is best-in-class but requires you to train the LoRA. Sloane's per-persona LoRA is the frictionless version — same face across every generation, out of the box.
Free tier realities
What each product actually gives you on free tier (verified 2026-09):
Sloane free: 50 messages/day, 5 free custom photo generations at signup, unlimited cadence photos (persona sends them proactively), any of the ~80 personas. Roleplay skill requires Plus ($9.99).
Character.AI free: 50-80 messages/day (cap varies), full character library, no photo generation, character drift over long conversations.
SillyTavern free: Software is free, but you're paying either hardware or API costs. No product-level free tier — it's bring-your-own-everything.
Muah AI free: Limited trial (a few messages), then subscription required for real use.
Kindroid free: Limited trial. Meaningful use requires subscription.
Janitor AI free: Software is free, but you need an API key (from OpenRouter, DeepInfra, etc.) which costs money for real usage.
Sloane's free tier is the closest to "usable indefinitely without paying" — 50 messages/day with photo cadence is enough for daily-use roleplay if you don't need custom photos beyond the 5 grant.
Pick based on what you actually want
"I want structured roleplay with photos at beats" → Sloane (Roleplay skill on Plus tier).
"I want maximum control and I'm technical" → SillyTavern + local LLM.
"I want a specific niche character someone else already wrote" → Character.AI (biggest catalog).
"I want voice + roleplay together" → Muah AI (accept the coin economy).
"I want character depth over photo integration" → Kindroid.
"I want unfiltered and I already have API access" → Janitor AI.
For most users looking for "AI roleplay chatbot that works out of the box with photo delivery," Sloane's Roleplay skill is the shortest path. Free tier is enough to check whether the shape of the product fits.