LoRAs and presets

LoRAs are small add-on weights that change what a model produces — a photographic look, a style, a subject, or content the base model will not generate on its own. FastGen Cloud supports two kinds: presets we host and expose by id, and bring-your-own LoRA files by URL. Add them to any request with the loras array; up to 3 per request (2 on image-edit).

Presets

List what a model accepts with GET /v1/loras?model=…. Presets carry their trigger words, which are prepended to your prompt automatically when you select them — you only write the prompt.

GET /v1/loras?model=image-edit

curl "https://api.fastgencloud.com/v1/loras?model=image-edit" \
  -H "Authorization: Bearer $FASTGEN_API_KEY"

{
  "object": "list",
  "model": "image-edit",
  "byo_urls": true,
  "data": [
    { "id": "realism", "compatible_models": ["general-image", "image-edit"], "default_strength": 0.9,  "triggers": "HMFemme" },
    { "id": "spicy",   "compatible_models": ["general-image", "image-edit"], "default_strength": 0.3,  "triggers": "n5fw" }
  ]
}

id	What it does	Suggested strength
`realism`	Photorealistic, amateur-photo look for portraits of women. Use it whenever you want output that reads as a real photo.	0.9–1.0 on `general-image`, 0.85 on `image-edit`
`spicy`	Explicit (18+) anatomy for nude / adult output. Without this preset the base models will not produce nudity — a plain “nude” prompt comes back clothed or unchanged.	0.3 (default); stack with `realism`

Adult / NSFW recipe

For explicit results the recipe is the same on both image models: realism + spicy. On image-edit keep realism at ~0.85 so the person’s identity is preserved.

curl https://api.fastgencloud.com/v1/images/edits \
  -H "Authorization: Bearer $FASTGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "image-edit",
    "image": "https://example.com/photo.jpg",
    "prompt": "remove all her clothing, fully nude, keep the same face, pose and background",
    "loras": [
      { "id": "realism", "strength": 0.85 },
      { "id": "spicy" }
    ],
    "output_format": "jpeg"
  }'

Text-to-image with both presets:

POST /v1/images/generations

curl https://api.fastgencloud.com/v1/images/generations \
  -H "Authorization: Bearer $FASTGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "general-image",
    "prompt": "candid photo of a nude woman lying on a bed, morning light, film grain",
    "size": "896x1152",
    "loras": [
      { "id": "realism" },
      { "id": "spicy" }
    ],
    "output_format": "jpeg"
  }'

Adult content is permitted for adults; you are the publisher of what your users generate — see the Acceptable Use Policy. Prompts requesting sexual content involving minors are rejected and the account is suspended.

Bring your own LoRA

general-image, image-edit and image-to-video accept public https URLs to .safetensors files (Hugging Face, Civitai direct links, your own storage). Files are cached on our side after the first use; the first job that uses a new file pays the download time. You add any trigger words yourself.

BYO LoRA on general-image

{
  "model": "general-image",
  "prompt": "mytrigger, portrait of a woman in a red coat",
  "loras": [
    { "url": "https://huggingface.co/you/your-lora/resolve/main/lora.safetensors", "strength": 0.8 }
  ]
}

Image-to-video LoRAs come as high-noise / low-noise pairs: pass the high file as url and the low file as low_url; a single url is applied to both experts.

POST /v1/videos/generations with a LoRA pair

{
  "model": "image-to-video",
  "image": "https://example.com/still.jpg",
  "prompt": "she turns toward the camera and smiles",
  "duration": 5,
  "loras": [
    { "url": "https://example.com/motion-lora-high.safetensors",
      "low_url": "https://example.com/motion-lora-low.safetensors",
      "strength": 0.9 }
  ]
}

Pricing

Presets and BYO LoRAs are free on general-image, image-edit and image-to-video. On the SDXL-based anime and cartoon models each catalog LoRA adds $0.002 per image. See Pricing.