Prompt String: MKAI_TaliaYangCER, dark skinned female, pink hair, slit pupils, blue eyes, hair intakes, huge ahoge, medium hair, single braid, very long braid, robot joints, prosthetic legs, prosthetic arms, mouth mask, black jumpsuit, spiked footwear, black jacket,
Recommendation for users: I focused this data set to include as many reference to her legs as i could its not perfect. Increase the weight of her prosthetic arms and legs (try 1.1–1.3). Adding black jacket to your prompt also improves outfit accuracy.
Character LoRA for Talia Yang from Cyberpunk: Edgerunners.
Trigger Words
Main trigger: MKAI_TaliaYangCER
Appearance: dark skinned female, pink hair, slit pupils, blue eyes, hair intakes, huge ahoge, medium hair, single braid, very long braid
Cyberware: robot joints, prosthetic legs, prosthetic arms, mouth mask
Outfit: black jumpsuit, spiked footwear, black jacket
Transparent mask:
```text
MKAI_TaliaYangCER,dark skinned female,pink hair,slit pupils,blue eyes,hair intakes,huge ahoge,medium hair,single braid,very long braid,robot joints,prosthetic legs, prosthetic arms,transparent mask,,black jumpsuit,spiked footwear,
Description
Character LoRA for Talia Yang, focused on her dark skin, pink hair, prominent ahoge and long braid, blue eyes with slit pupils, and mechanical limbs. Her design includes a black jumpsuit and spiked footwear, with mouth-mask and transparent-mask prompt variants.
For the transparent-mask variant, use a close-up and explicitly describe the mouth or smile being visible through the mask.
MKAI_TaliaYangCER,dark skinned female,pink hair,slit pupils,blue eyes,hair intakes,huge ahoge,medium hair,single braid,very long braid,robot joints,prosthetic legs, prosthetic arms,mouth mask,black jumpsuit,spiked footwear
Transparent-mask face
MKAI_TaliaYangCER,dark skinned female,pink hair,slit pupils,blue eyes,hair intakes,huge ahoge,medium hair,single braid,very long braid,robot joints,prosthetic legs, prosthetic arms,transparent mask,black jumpsuit,spiked footwear,
Use Detailers for better face and eye detail!
Training Parameter - Let me know what where I can improve!
Setting | Value |
|---|---|
| Trainer | AI Toolkit 0.13.34 |
| Base model | Anima Base v1.0 (original Anima) |
| Dataset | 49 images with paired captions |
| Training steps | 2,000 |
| Repeats | 1 |
| Rank / alpha | 16 / 16 |
| Optimizer | AdamW 8-bit |
| Learning rate | 0.0001 / 1e-4 |
| LR scheduler | Constant |
| Weight decay | 0.0001 |
| Batch / gradient accumulation | 1 / 1 |
| Training resolution | 1024-area aspect-ratio buckets |
| Horizontal flipping | Enabled |
| Caption dropout | 5% |
| Compute precision | BF16 |
| Saved adapter precision | BF16 |
| Text encoder training | Disabled |
| EMA | Disabled |
| Timestep sampling | Sigmoid |
| Checkpoint interval | 250 steps