Same girl, different morning

Solo-character imagery of your companion across the three model families. The range goes from clothed portrait shots and casual lifestyle frames through fully explicit NSFW output for verified 18+ accounts. Pose, outfit, lighting, location, and mood all read from natural-language prompts. Illegal categories are blocked at the model level no matter how the prompt is phrased; everything legal that adults want to render is in scope.

The free tier ships with a thirty-per-day quota across all three model families, NSFW included. Premium drops the cap entirely, lifts the resolution ceiling, gives you priority during peak load, and stops counting upscales and regenerations against your daily total. Casual users almost never run into the thirty-image limit; the cap is more of a soft constraint against gallery-mining workloads than a teaser meant to push you onto premium.

Three live pipelines. Realistic targets photographic output with natural skin tone and plausible lighting. Anime renders drawn-line panels while preserving the character anchor across frames. The 2.5D pipeline lives between the other two — realistic proportions, stylized features, softened edges. The same character moves between all three without losing recognizability, because the identity vector is stored once and read by every model.

That's the headline feature, not a workaround. The identity vector locks on the first successful generation and every subsequent prompt references it. Change the location, the outfit, the lighting, the model family — the face and body still read as the same person. The reason this works is that the pipeline does not re-derive identity from the prompt every request, which is the specific failure mode behind drift in most competing generators.

Adult content is a baseline capability on the free tier, gated only by the standard daily quota. Trynectar AI runs as an 18+ product end-to-end — there is no SFW mode toggle you flip on, no series of clicks to "unlock" the adult surface. Illegal categories are blocked at the model level on every tier with no exceptions; everything legal that an adult would want to render is in scope from the first generation onward.

Free tier renders at 1024×1024 native and offers a 1.5x upscale to 1536×1536. Premium lifts native to 2048×2048 and the upscale path to 4096×4096. The upscaler is a generative pass — a second model adds plausible detail rather than smearing pixels — which is why character consistency carries through the upscale instead of breaking on it. Bicubic filtering it is not.

Long descriptive prompts are the path to good output. The parser was trained on natural-language descriptions, not keyword strings, so light direction, hour of the day, specific wardrobe items, exact posture, props, and framing all read as honored constraints. The pipeline doesn't smooth your input back down to a generic body-on-bed compromise the way most open-source generators do; the more specific the prompt, the more specific the rendered frame.

Every output is private by default — encrypted at rest, account-scoped, never visible to other users, never pulled into training data, never handed to third parties. Per-image deletion is one click in the gallery view. A full gallery clear is one click in settings. The account-level wipe purges everything including the memory graph behind the chat. The privacy posture is identical on free and premium; it's not a paid feature.