<h1 id="sebastian-+-jessica-v2-dual-character-lora-(microsoft-lens)">Sebastian + Jessica v2 — Dual Character LoRA (Microsoft Lens)</h1>This is the first Microsoft Lens LoRA ever generated - considering we JUST created the first trainer ever to exist in the world. You can check it out in the links below. Why Microsoft Lens? Because it's tiny but mighty. Super small model size but super high quality images. Not only does the model run on our dev machines but also the training end to end on an Nvidia RTX 5060ti with 16GB of VRAM. This is NSFW sample lora created to show case that the LoRA training works. If you try it out let us know! LoRA fine-tune for Microsoft Lens — two recurring characters:<ul><li>Sebastian — blonde man, long hair, Irish blue eyes, tuxedo</li><li>Jessica — redhead, long copper-red hair, Irish green eyes, purple dress</li></ul>Trained on Lens-Base (<a target="_blank" rel="ugc" href="https://huggingface.co/microsoft/Lens-Base">Hugging Face</a>) with <a target="_blank" rel="ugc" href="https://github.com/LoboForge/LoboForge-LensTrainer">LoboForge-LensTrainer</a>.Also on Hugging Face: <a target="_blank" rel="ugc" href="https://huggingface.co/LoboForge/lens-lora-sebastian-jessica-v2">LoboForge/lens-lora-sebastian-jessica-v2</a><h2 id="training">Training</h2>| Field | Value | |-------|-------| | Dataset | 24 image/caption pairs, 1024×1024 | | Steps | 8000 | | LoRA rank / alpha | 16 / 16 | | Optimizer | AdamW 8-bit |<h2 id="prompt-format">Prompt format</h2>Full-sentence captions — no single trigger token. Describe both characters and the scene, matching your training captions.Example:<code>Sebastian is a blonde man with long hair and irish blue eyes in a tuxedo. Jessica is a redhead woman with long copper colored red hair and irish green eyes in a purple dress named. They are standing together facing forward and laughing</code><h2 id="inference">Inference</h2>| Setting | Value | |---------|-------| | Base | Lens or Lens-Base (HF) | | Steps | 50 | | CFG | 5.0 | | Resolution | 1024×1024 |ComfyUI: load Lens checkpoint + this LoRA (<code>diffusion_model.*</code> keys).<h2 id="license">License</h2>PolyForm Noncommercial License 1.0.0 — non-commercial use only.

Micoroft Lens (World First) - Sebastian + Jessica v2 — Dual Character LoRA

Sebastian + Jessica v2 — Dual Character LoRA (Microsoft Lens)

LoRA fine-tune for Microsoft Lens — two recurring characters:

<ul>
<li>Sebastian — blonde man, long hair, Irish blue eyes, tuxedo</li>
<li>Jessica — redhead, long copper-red hair, Irish green eyes, purple dress</li>
</ul>

Trained on Lens-Base (<a href="https://huggingface.co/microsoft/Lens-Base" rel="ugc">Hugging Face</a>) with <a href="https://github.com/LoboForge/LoboForge-LensTrainer" rel="ugc">LoboForge-LensTrainer</a>.

Also on Hugging Face: <a href="https://huggingface.co/LoboForge/lens-lora-sebastian-jessica-v2" rel="ugc">LoboForge/lens-lora-sebastian-jessica-v2</a>

Training

| Field | Value |
|-------|-------|
| Dataset | 24 image/caption pairs, 1024×1024 |
| Steps | 8000 |
| LoRA rank / alpha | 16 / 16 |
| Optimizer | AdamW 8-bit |

Prompt format

Full-sentence captions — no single trigger token. Describe both characters and the scene, matching your training captions.

Example:

<code>
Sebastian is a blonde man with long hair and irish blue eyes in a tuxedo. Jessica is a redhead woman with long copper colored red hair and irish green eyes in a purple dress named. They are standing together facing forward and laughing
</code>

Inference

| Setting | Value |
|---------|-------|
| Base | Lens or Lens-Base (HF) |
| Steps | 50 |
| CFG | 5.0 |
| Resolution | 1024×1024 |

ComfyUI: load Lens checkpoint + this LoRA (<code>diffusion_model.*</code> keys).

License

PolyForm Noncommercial License 1.0.0 — non-commercial use only.

Files with SHA256: 505f45228707620011f2b20c0b29a4798b103e685870e4d9fd77e620953b88e5