# Stable Diffusion with LoRAs API

> Text-to-image with any Stable Diffusion base model plus LoRAs, embeddings, ControlNets and IP adapters

Generates images from a text prompt using a base model given by URL or HuggingFace ID in `model_name`. Several `loras` can be merged, and embeddings, ControlNets, IP adapters and IC Light inputs are supported. Sampler options include many schedulers and steps from 1 to 150. Generates 1 to 8 images in png or jpeg.

- Model id: `lora`
- Made by: Open model
- Category: Text to Image
- Output: image
- Estimated price per run: $0.057 / image (cheapest run $0.057 (57 API credits))
- Run: `POST https://api.doitong.com/v1/run/lora`
- Quote: `POST https://api.doitong.com/v1/quote/lora`
- Docs page: https://doitong.com/en-US/developers/models/lora

## Pricing

Price: **$0.057 / image**. Estimated from the typical cost of one run, per output (`num_images`). The quote is exactly what the run charges.

The full example input below costs $0.057 = 57 API credits (1 image × 57 API credits).

Prices are the upstream price in USD, paid from the API balance in API credits (1000 API credits = 1 USD). `POST https://api.doitong.com/v1/quote/lora` with the same JSON body returns the exact price for that input; the quote is exactly what a run charges. Invalid input is never charged and failed or canceled requests are refunded automatically.

## Input parameters

Send either the canonical name or the native field name (not both). Canonical names are the same across models.

| Parameter | Canonical name | Type | Required | Default | Allowed values / range | Description |
|---|---|---|---|---|---|---|
| `model_name` |  | string | yes |  |  | Base model for generating the image, identified by either a URL or a HuggingFace ID. |
| `unet_name` |  | string \| null | no |  |  | Custom U-Net selection for generating images; supply either its HuggingFace ID or its URL. |
| `variant` |  | string \| null | no |  |  | HuggingFace model variant selection; 'fp16' is one example. |
| `prompt` |  | string | yes |  |  | Describe the image you want generated, using as much detail as possible to get the best results. |
| `negative_prompt` |  | string | no | `` |  | Negative prompt specifying unwanted image details. It can exclude colors, objects, scenery, or finer features and qualities, such as a moustache, blur, or low resolution. |
| `prompt_weighting` |  | boolean | no | `false` |  | Prompt-weighting toggle: true removes the 77 token limit through embedding averaging and enables weighting syntax. |
| `loras` |  | array of object {path, scale} | no | `[]` |  | LoRAs whose combined effect produces the generated image, with no limit on how many you supply. |
| `loras[].path` |  | string | yes |  |  | Location of the LoRA weights, specified as a URL or path. |
| `loras[].scale` |  | number | no | `1` | 0 – 4 | Multiplier applied to the LoRA weight before it is merged into the base model. |
| `embeddings` |  | array of object {path, tokens} | no | `[]` |  | Embeddings applied during image generation to map prompt tokens to embedding weights. Currently, only one embedding is supported. |
| `embeddings[].path` |  | string | yes |  |  | Embedding weights specified by URL or file path. |
| `embeddings[].tokens` |  | array of string \| null | no |  |  | Prompt tokens mapped to embedding weights. Omitting them creates one token per embedding-file vector, named <s0>, <s1>, ...; use the mapped tokens in your prompts. |
| `controlnets` |  | array of object {path, config_url, variant, image_url, mask_url, conditioning_scale, start_percentage, end_percentage, ip_adapter_index} | no | `[]` |  | Control nets scheduled for application to the image at specified timesteps during generation, with no limit on how many you supply. |
| `controlnets[].path` |  | string | yes |  |  | ControlNet weight location, specified as a URL or path. |
| `controlnets[].config_url` |  | string \| null | no |  |  | Optional URL pointing to the ControlNet's config.json file. |
| `controlnets[].variant` |  | string \| null | no |  |  | Optional variant for use with a Hugging Face repository key. |
| `controlnets[].image_url` |  | string (image URL) | yes |  |  | Image URL supplied as the ControlNet input. |
| `controlnets[].mask_url` |  | string (image URL) \| null | no |  |  | ControlNet mask, which must match the control image in size when supplied; width and height must each be divisible by 32. |
| `controlnets[].conditioning_scale` |  | number | no | `1` | 0 – 2 | Scaling applied to ControlNet weights before they are merged into the base model. |
| `controlnets[].start_percentage` |  | number | no | `0` | 0 – 1 | Percentage of total timesteps at which ControlNet starts influencing image generation. |
| `controlnets[].end_percentage` |  | number | no | `1` | 0 – 1 | Percentage of total timesteps at which ControlNet stops influencing image generation. |
| `controlnets[].ip_adapter_index` |  | integer \| null | no |  |  | IP adapter selection by index for an InstantID ControlNet; other ControlNets do not need this setting. |
| `controlnet_guess_mode` |  | boolean | no | `false` |  | Restricts ControlNet application to conditional predictions when true. |
| `ip_adapter` |  | array of object {ip_adapter_image_url, ip_adapter_mask_url, path, model_subfolder, weight_name, insight_face_model_path, scale, scale_json, unconditional_noising_factor, image_projection_shortcut} | no | `[]` |  | IP adapters used to generate images. Providing any entries requires image_encoder_path. |
| `ip_adapter[].ip_adapter_image_url` |  | string (image URL) \| array of image URL | yes |  |  | Image URL supplied as the IP adapter input. |
| `ip_adapter[].ip_adapter_mask_url` |  | string (image URL) \| null | no |  |  | IP adapter mask, subject to a size constraint when supplied: its dimensions must equal those of the adapter image. |
| `ip_adapter[].path` |  | string | yes |  |  | IP adapter weights location, specified as a URL or path. |
| `ip_adapter[].model_subfolder` |  | string \| null | no |  |  | IP adapter weight location, specified as a subfolder within the model directory. |
| `ip_adapter[].weight_name` |  | string \| null | no |  |  | Filename of the weights. |
| `ip_adapter[].insight_face_model_path` |  | string \| null | no |  |  | InsightFace model weights location, specified as a URL or path. |
| `ip_adapter[].scale` |  | number | no | `1` | ≥ 0 | Scaling factor for IP adapter weights, applied before they are merged into the base model. |
| `ip_adapter[].scale_json` |  | object \| null | no |  |  | Scaling factor for IP adapter weights, applied before they are merged into the base model. |
| `ip_adapter[].unconditional_noising_factor` |  | number | no | `0` | 0 – 1 | Factor controlling unconditional noising for the IP adapter. |
| `ip_adapter[].image_projection_shortcut` |  | boolean | no | `true` |  | Image projection shortcut setting, with True as the default. Use False for FaceID plus V1 models and True for FaceID plus V2 models. |
| `image_encoder_path` |  | string \| null | no |  |  | Image encoder specified by an HTTPS URL or Hugging Face ID, with compatibility determined by the IP-Adapter weights: SD1.5 uses image_encoder_subfolder='models/image_encoder', while h94/IP-Adapter SDXL uses h94/IP-Adapter and image_encoder_subfolder='sdxl_models/image_encoder'. An encoder is required whenever ip_adapter contains a value. |
| `image_encoder_subfolder` |  | string \| null | no |  |  | Subdirectory within image_encoder_path that holds the image encoder weights. |
| `image_encoder_weight_name` |  | string \| null | no | `pytorch_model.bin` |  | Select the image encoder's weights by name for use in generating images. |
| `ic_light_model_url` |  | string \| null | no |  |  | IC Light model URL used when generating images. |
| `ic_light_model_background_image_url` |  | string (image URL) \| null | no |  |  | Model-compatible background for IC Light image generation, supplied as an image URL. |
| `ic_light_image_url` |  | string (image URL) \| null | no |  |  | Input image URL for generation with the IC Light model. |
| `seed` |  | integer \| null | no |  |  | Seed for reproducing images in Stable Diffusion. An image can be reproduced exactly on every run by keeping the model version, prompt, and seed unchanged. |
| `image_size` |  | object {width, height} \| enum (string) \| null | no | `square_hd` | `square_hd`, `square`, `portrait_4_3`, `portrait_16_9`, `landscape_4_3`, `landscape_16_9` | Output image dimensions, selected from presets or specified as custom width and height. Both custom dimensions must be divisible by 8. |
| `num_inference_steps` |  | integer | no | `30` | 1 – 150 | Control how many generation steps Stable Diffusion performs; additional steps may yield a more detailed final image. |
| `guidance_scale` |  | number | no | `7.5` | 0 – 20 | Prompt-adherence strength for image selection, expressed as CFG (Classifier Free Guidance). |
| `clip_skip` |  | integer | no | `0` | 0 – 2 | Skips a portion of image generation to render faster, producing slightly different results. |
| `scheduler` |  | enum (string) \| null | no |  | `DPM++ 2M`, `DPM++ 2M Karras`, `DPM++ 2M SDE`, `DPM++ 2M SDE Karras`, `Euler`, `Euler A`, `Euler (trailing timesteps)`, `LCM`, `LCM (trailing timesteps)`, `DDIM`, `TCD` | Scheduler or sampler used to denoise the image. |
| `timesteps` |  | object {method, array} | no | `{"array":[],"method":"default"}` |  | Denoising timestep schedule override, mutually exclusive with sigmas. Without an override—the default—the scheduler calculates timesteps from num_inference_steps; supplying a custom schedule makes num_inference_steps irrelevant. An override requires scheduler support for the timesteps argument to set_timesteps. |
| `timesteps.method` |  | enum (string) | no | `default` | `default`, `array` | Choose how timesteps are determined: 'default', the default setting, lets the scheduler use `num_inference_steps`. To supply a timestep schedule yourself, select 'array' and place the schedule in the `array` field. |
| `timesteps.array` |  | array of integer | no | `[]` |  | Timestep schedule applied when the selected method is 'custom'. |
| `sigmas` |  | object {method, array} | no | `{"array":[],"method":"default"}` |  | Denoising sigma schedule override, mutually exclusive with timesteps. Without an override—the default—the scheduler calculates sigmas from num_inference_steps; supplying a custom schedule makes num_inference_steps irrelevant. An override requires scheduler support for the sigmas argument to set_sigmas. |
| `sigmas.method` |  | enum (string) | no | `default` | `default`, `array` | Choose how sigmas are determined: 'default', the default setting, uses the scheduler's own sigma values. To supply a sigma schedule yourself, select 'custom' and place the schedule in the `array` field. |
| `sigmas.array` |  | array of number | no | `[]` |  | Sigma schedule applied when the selected method is 'custom'. |
| `prediction_type` |  | enum (string) | no | `epsilon` | `v_prediction`, `epsilon` | Choose the prediction type for generating images; without a different selection, `epsilon` is used. |
| `rescale_betas_snr_zero` |  | boolean | no | `false` |  | Controls whether the sampler enables rescale_betas_snr_zero. |
| `image_format` |  | enum (string) | no | `png` | `jpeg`, `png` | File format for the resulting image. |
| `num_images` | `num_outputs` | integer | no | `1` | 1 – 8 | Set how many images a single request produces, allowing more generation time for a larger batch. |
| `enable_safety_checker` |  | boolean | no | `false` |  | Safety checker activation; true enables checking. |
| `tile_width` |  | integer | no | `4096` | 128 – 4,096 | Set how large each tile is during image generation. |
| `tile_height` |  | integer | no | `4096` | 128 – 4,096 | Set how large each tile is during image generation. |
| `tile_stride_width` |  | integer | no | `2048` | 64 – 2,048 | Set the spacing between successive tile positions during image generation. |
| `tile_stride_height` |  | integer | no | `2048` | 64 – 2,048 | Set the spacing between successive tile positions during image generation. |
| `eta` |  | number | no | `0` | 0 – 1 | Eta setting used when generating images. |
| `debug_latents` |  | boolean | no | `false` |  | Saves latents for debugging when true. |
| `debug_per_pass_latents` |  | boolean | no | `false` |  | Enable per-pass debugging by setting this to true, which saves the latents from every pass. |

File parameters (`ic_light_model_background_image_url`, `ic_light_image_url`) take a public `https://` URL, a `data:` URI up to 10 MB, or the `file_url` of an upload from `POST /files`.

## Example input

Required parameters only:

```json
{
  "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
}
```

A fuller example:

```json
{
  "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
  "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
}
```

## Run it

Set `DOITONG_API_KEY` to a key from https://doitong.com/developers/dashboard#keys. `?wait=20` returns the finished request in the same call when it is ready within 20 seconds; otherwise poll `GET /requests/{id}` (it also accepts `?wait=`), or pass `webhook_url`.

### curl

```bash
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/lora?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
  }'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```

### JavaScript

```javascript
const API = "https://api.doitong.com/v1";
const headers = {
  "Authorization": "Bearer " + process.env.DOITONG_API_KEY,
  "Content-Type": "application/json",
};

// Run the model; wait up to 20 s for the finished result in the same call
let res = await fetch(API + "/run/lora?wait=20", {
  method: "POST",
  headers,
  body: JSON.stringify({
    "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
  }),
});
let request = await res.json();
if (!res.ok) throw new Error(request.error.code + ": " + request.error.message);

// Not finished yet: long-poll until it is
while (request.status === "queued" || request.status === "running") {
  res = await fetch(API + "/requests/" + request.id + "?wait=20", { headers });
  request = await res.json();
}
if (request.status !== "succeeded") throw new Error(JSON.stringify(request.error));
console.log(request.outputs.map((o) => o.url), request.cost.credits + " API credits ($" + request.cost.usd + ")");
```

### Python

```python
import os
import requests

API = "https://api.doitong.com/v1"
headers = {"Authorization": "Bearer " + os.environ["DOITONG_API_KEY"]}

payload = {
    "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field",
}

# Run the model; wait up to 20 s for the finished result in the same call
r = requests.post(API + "/run/lora", params={"wait": 20}, json=payload, headers=headers)
request = r.json()
if not r.ok:
    raise RuntimeError(request["error"]["code"] + ": " + request["error"]["message"])

# Not finished yet: long-poll until it is
while request["status"] in ("queued", "running"):
    request = requests.get(API + "/requests/" + request["id"], params={"wait": 20}, headers=headers).json()
if request["status"] != "succeeded":
    raise RuntimeError(request["error"])
print([o["url"] for o in request["outputs"]], request["cost"]["credits"], "API credits", "$" + request["cost"]["usd"])
```

### MCP

The MCP server at `https://api.doitong.com/v1/mcp` exposes the tools `search_models`, `get_model`, `quote`, `run`, `get_request`, `cancel_request`, `list_requests`, `create_upload` and `get_account`. A `run` call for this model:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "run",
    "arguments": {
      "model_id": "lora",
      "input": {
        "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
        "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
      },
      "wait_seconds": 20
    }
  }
}
```

## Response

Every call returns a request object. `status` is one of `queued`, `running`, `succeeded`, `failed`, `canceled`. `outputs` lists the generated files with URLs on our storage (kept forever, or 7 days with `?storage=temporary`; `storage` says which); `output` is the model's own result object with the same URLs.

```json
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "lora",
  "status": "succeeded",
  "created_at": "2026-10-20T12:00:00Z",
  "started_at": "2026-10-20T12:00:02Z",
  "completed_at": "2026-10-20T12:00:09Z",
  "queue_position": null,
  "cost": {
    "credits": 57,
    "usd": "0.057",
    "charged": 57,
    "refunded": 0
  },
  "input": {
    "model_name": "stabilityai/stable-diffusion-xl-base-1.0",
    "prompt": "A red fox trots through fresh snow at sunrise, soft golden light, shallow depth of field"
  },
  "output": {
    "images": [
      {
        "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
        "content_type": "image/jpeg",
        "file_name": "0-output.jpeg",
        "file_size": 1843200,
        "width": 1024,
        "height": 768
      }
    ],
    "seed": 42,
    "has_nsfw_concepts": [
      false
    ],
    "debug_latents": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg",
      "file_name": "0-output.jpeg",
      "file_size": 1843200
    },
    "debug_per_pass_latents": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg",
      "file_name": "0-output.jpeg",
      "file_size": 1843200
    }
  },
  "outputs": [
    {
      "type": "image",
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg",
      "file_size": 1843200,
      "width": 1024,
      "height": 768,
      "stored": true
    }
  ],
  "error": null,
  "metrics": {
    "inference_seconds": 6.8
  },
  "urls": {
    "self": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t",
    "cancel": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel"
  }
}
```

### Output fields (`output`)

| Field | Type | Description |
|---|---|---|
| `images` | array of object {url, content_type, file_name, file_size, width, height} | File information for the generated images. |
| `seed` | integer | Image-generation seed, including the randomly chosen value if the input omitted a seed; otherwise, it repeats the supplied value. |
| `has_nsfw_concepts` | array of boolean | Indicates whether NSFW concepts appear in the generated images. |
| `debug_latents` | object {url, content_type, file_name, file_size} \| null | Latent data stored for debugging. |
| `debug_per_pass_latents` | object {url, content_type, file_name, file_size} \| null | Per-pass latent data stored for debugging. |

## Related models

- [Stable Diffusion with LoRAs · Image to Image](https://doitong.com/developers/models/lora/image-to-image.md) (`lora/image-to-image`): $0.264 / image
- [Stable Diffusion with LoRAs · Inpaint](https://doitong.com/developers/models/lora/inpaint.md) (`lora/inpaint`): $0.184 / image
- [FLUX.1 \[dev\]](https://doitong.com/developers/models/flux/dev.md) (`flux/dev`): $0.025 / megapixel
- [FLUX.1 \[schnell\]](https://doitong.com/developers/models/flux/schnell.md) (`flux/schnell`): $0.003 / megapixel
- [FLUX.2 \[dev\]](https://doitong.com/developers/models/flux-2.md) (`flux-2`): $0.012 / megapixel
- [FLUX.2 \[pro\]](https://doitong.com/developers/models/flux-2-pro.md) (`flux-2-pro`): $0.015 / megapixel of output image
- [GPT Image 2.5 Text to Image](https://doitong.com/developers/models/openai/gpt-image-2.5/flare/text-to-image.md) (`openai/gpt-image-2.5/flare/text-to-image`): from $0.008 / image
- [Ideogram V4.5](https://doitong.com/developers/models/ideogram/v4.5.md) (`ideogram/v4.5`): from $0.03 / image

## Links

- HTML page: https://doitong.com/developers/models/lora
- All models and guides for AI agents: https://doitong.com/developers/llms.txt
- OpenAPI 3.1 with this model's typed request body: https://doitong.com/developers/openapi.json?model=lora
- Errors: https://doitong.com/developers/docs/errors
- Model details as JSON: https://api.doitong.com/v1/models/lora