---
title: DreamBoat
emoji: "⛵"
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: "5.31.0"
app_file: app.py
pinned: false
license: apache-2.0
tags:
  - flux
  - kontext
  - character-training
  - nsfw
  - lgbtq
---

# DreamBoat

Character training image generator for consistent AI characters. Companion to [PrideWAN](https://huggingface.co/spaces/Jds20001/PrideWAN).

Uses FLUX.1-Kontext to maintain character identity across varied poses, angles, lighting, and content levels. Generate 15-25 training images, download as a zip, and train your character LoRA with [ai-toolkit](https://github.com/ostris/ai-toolkit).

## The three tabs

1. **Norma Jean (Identity)** — FLUX.1-Kontext-dev. Upload one reference photo and
   generate identity-consistent variations (angles, lighting, outfits, shirtless).
   Kontext holds the face/body across scenes.
2. **Marilyn (Explicit)** — FLUX.1-dev img2img with NSFW LoRAs. Take a Norma Jean
   shot and transform it; FLUX.1-dev's lighter safety training lets the NSFW LoRAs work.
   Use the Character Description box + Transformation Strength 0.75-0.85 to keep the face.
3. **Street Pack (Batch)** — one reference photo in, a full set of candid
   social-media-style identity shots out, dropped into the gallery live.

A **Realism** LoRA runs on both pipelines and a **Candid** toggle appends an
iPhone/handheld/natural-grain suffix — both fight the plastic "obviously AI" FLUX look.

## Setup

This Space uses two **gated** base models. Before it will boot, accept terms for both
(one click each, with the same HF account whose token you add as a Space secret):

1. [black-forest-labs/FLUX.1-Kontext-dev](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev)
2. [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev)

Then add your HuggingFace token as a Space secret named **`HF_TOKEN`**.

The NSFW and realism LoRAs are public (no terms to accept):
`Heartsync/Flux-NSFW-uncensored`, `lustlyai/Flux_Lustly.ai_Uncensored_nsfw_v1`,
and `Shakker-Labs/FLUX.1-dev-LoRA-add-details`.

## How It Works

1. Upload one clear reference photo of your character
2. Use presets or write prompts describing the variation you want
3. Kontext maintains identity from the reference while generating the new pose/angle/scene
4. Collect 15-25 varied images into the training set
5. Download as a zip, ready for ai-toolkit LoRA training

## Training Set Tips

A good training set includes:
- 5 face close-ups (varied angles and lighting)
- 5 upper body / torso shots
- 5 full body shots (varied poses)
- 5+ explicit / anatomical shots (if training for NSFW content)
- Simple, solid backgrounds throughout

## Credits

- [FLUX.1-Kontext-dev](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) and [FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) by Black Forest Labs
- NSFW LoRAs by [Heartsync](https://huggingface.co/Heartsync/Flux-NSFW-uncensored) and [lustlyai](https://huggingface.co/lustlyai/Flux_Lustly.ai_Uncensored_nsfw_v1)
- Realism LoRA (add-details) by [Shakker-Labs](https://huggingface.co/Shakker-Labs/FLUX.1-dev-LoRA-add-details)
- Built for the [PrideWAN](https://huggingface.co/spaces/Jds20001/PrideWAN) ecosystem