Pruna AI

Access Pruna AI's optimized image generation models including P-Image and P-Image-Edit through Runware's unified API. Real-time performance with production-grade quality and efficiency.

Introduction

Pruna AI's models are integrated into the Runware platform through our unified API, providing access to highly optimized image generation technology designed for real-time production workflows. Built on Pruna's AI optimization framework, these models deliver performance-oriented generation with efficient resource utilization.

Through the providerSettings.prunaai object, you can access Pruna's unique features such as turbo mode acceleration, while maintaining the consistency of Runware's standard API structure. This page documents the technical specifications, parameter requirements, and provider-specific settings for all Pruna AI models available through our platform.

providerSettings » prunaai
prunaai
object

Configuration object for Pruna AI-specific image generation settings and optimization controls.

Example 1 example 
{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "positivePrompt": "A product photograph with clean composition",
  "model": "prunaai:2@1",
  "width": 1024,
  "height": 1024,
  "providerSettings": {
    "prunaai": {
      "turbo": true
    }
  }
}
Properties 1 property
providerSettings » prunaai » turbo
turbo
boolean Default: false

Enables additional performance optimizations for faster generation. When activated, the model applies aggressive acceleration techniques to reduce inference time.

For complex or detailed generation tasks, it is recommended to disable turbo mode to maintain optimal quality. Turbo mode prioritizes speed over subtle details.

Image models

P-Image

Real-time text-to-image model designed for instant, high-quality creation. Supports strong text rendering, precise prompt control, LoRA adaptations, and fast output suitable for production workflows with efficiency optimizations built in.

Model AIR ID: prunaai:1@1.

Supported workflows: Text-to-image.

Technical specifications:

  • Positive prompt: 1-2000 characters.
  • Supported dimensions: 512-2048 pixels (width and height, multiples of 16).
  • LoRA support: Single LoRA model via HuggingFace URL.
{
  "taskType": "imageInference",
  "taskUUID": "f47ac10b-58cc-4372-a567-0e02b2c3d496",
  "model": "prunaai:1@1",
  "positivePrompt": "A modern minimalist poster with bold typography reading 'INNOVATION' in clean sans-serif font",
  "width": 1024,
  "height": 1024
}
{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da7",
  "model": "prunaai:1@1",
  "positivePrompt": "A stylized character portrait with distinctive artistic aesthetic",
  "width": 1024,
  "height": 1024,
  "lora": {
    "model": "huggingface.co/username/model-name/lora-weights.safetensors",
    "weight": 1.5
  }
}

P-Image-Edit

Real-time editing model for multi-image refinement, layout adjustments, and controlled transformation with strong adherence to detailed instructions. Supports LoRA adaptations for style-specific editing workflows requiring precise modifications.

Model AIR ID: prunaai:2@1.

Supported workflows: Image-to-image.

Technical specifications:

  • Positive prompt: 1-2000 characters.
  • Reference images: Supports up to 4 images via inputs.referenceImages.
  • Supported dimensions: 1024×1024 (1:1), 1184×896 (4:3), 896×1184 (3:4), 1376×768 (16:9), 768×1376 (9:16), 1248×832 (3:2), 832×1248 (2:3).
  • LoRA support: Single LoRA model via HuggingFace URL.

When using reference images, dimensions are automatically matched to the input image if width and height are not specified.

Provider-specific settings:

Parameters supported: turbo.

{
  "taskType": "imageInference",
  "taskUUID": "f47ac10b-58cc-4372-a567-0e02b2c3d497",
  "model": "prunaai:2@1",
  "inputs": {
    "referenceImages": ["c64351d5-4c59-42f7-95e1-eace013eddab"]
  },
  "positivePrompt": "Change the background to a sunset beach scene while keeping the subject unchanged",
  "width": 1024,
  "height": 1024
}
{
  "taskType": "imageInference",
  "taskUUID": "6ba7b836-9dad-11d1-80b4-00c04fd430c8",
  "model": "prunaai:2@1",
  "inputs": {
    "referenceImages": [
      "c64351d5-4c59-42f7-95e1-eace013eddab",
      "d7e8f9a0-2b5c-4e7f-a1d3-9c8b7a6e5d4f"
    ]
  },
  "positivePrompt": "Combine these two images into a cohesive composition with unified lighting",
  "width": 1376,
  "height": 768
}
{
  "taskType": "imageInference",
  "taskUUID": "b8c4d952-7f27-4a6e-bc9a-83f01d1c6d63",
  "model": "prunaai:2@1",
  "inputs": {
    "referenceImages": ["c64351d5-4c59-42f7-95e1-eace013eddab"]
  },
  "positivePrompt": "Apply artistic style transformation while maintaining subject integrity",
  "lora": {
    "model": "huggingface.co/username/style-lora/weights.safetensors",
    "weight": 2.0
  }
}
{
  "taskType": "imageInference",
  "taskUUID": "550e8400-e29b-41d4-a716-446655440011",
  "model": "prunaai:2@1",
  "inputs": {
    "referenceImages": ["c64351d5-4c59-42f7-95e1-eace013eddab"]
  },
  "positivePrompt": "Adjust color temperature to warmer tones",
  "providerSettings": {
    "prunaai": {
      "turbo": true
    }
  }
}

Video models

P-Video

Pruna AI's P-Video is a real-time AI video generation model designed for fast creative iteration and production workflows. It supports text-to-video, image-to-video, and audio-to-video through a unified endpoint, delivering up to 1080p at 48 FPS with integrated dialogue generation and audio import. The model emphasizes speed, cost efficiency, and stable subject identity with a draft-to-refine pipeline that enables rapid iteration before final production.

Model AIR ID: prunaai:p-video@0.

Supported workflows: Text-to-video, image-to-video, audio-to-video.

Technical specifications:

  • Positive prompt: 1-2000 characters (required).
  • Supported dimensions:
    • 720p: 1280×720 (16:9), 720×1280 (9:16), 960×720 (4:3), 720×960 (3:4), 1080×720 (3:2), 720×1080 (2:3), 720×720 (1:1).
    • 1080p: 1920×1080 (16:9), 1080×1920 (9:16), 1440×1080 (4:3), 1080×1440 (3:4), 1620×1080 (3:2), 1080×1620 (2:3), 1080×1080 (1:1).
  • Dimension behavior:
    • Text-to-video: Specify explicit width and height from supported dimensions.
    • Image-to-video: Use resolution parameter (720p or 1080p) to automatically match aspect ratio from the input image.
  • Frame rate: 24 FPS or 48 FPS (default: 24).
  • Duration: 1-10 seconds (default: 5). Cannot be used when inputs.audio is provided (video duration automatically matches audio length).
  • Frame images: Supports first frame via inputs.frameImages.
  • Audio input: Supports audio conditioning via inputs.audio (maximum 60 seconds, cropped to 15 seconds).
  • Audio output: Generated by default, controlled via settings.audio parameter (adds ~1 second inference overhead when enabled).

Provider-specific settings:

  • settings.draft (boolean, default: false): Enable draft mode for lower-resolution preview generation.
  • settings.audio (boolean, default: true): Include audio in the generated video output.
  • settings.promptUpsampling (boolean, default: true): Automatically enhance the prompt for improved results.

Draft mode workflow: Enable draft: true with a specific seed to generate fast, lower-resolution previews (360p for 720p requests, 540p for 1080p requests). Once satisfied with the draft composition, run the same request with draft: false and identical parameters (seed, prompt, resolution, duration, fps) to produce the full high-quality version. The draft and full generation will match in color and motion, though the full version fixes artifacts and glitches present in the draft.

{
  "taskType": "videoInference",
  "taskUUID": "f47ac10b-58cc-4372-a567-0e02b2c3d498",
  "model": "prunaai:p-video@0",
  "positivePrompt": "A cinematic drone shot flying over a modern cityscape at golden hour",
  "width": 1920,
  "height": 1080,
  "duration": 8,
  "fps": 48,
  "settings": {
    "promptUpsampling": true,
    "audio": true
  }
}
{
  "taskType": "videoInference",
  "taskUUID": "6ba7b837-9dad-11d1-80b4-00c04fd430c8",
  "model": "prunaai:p-video@0",
  "inputs": {
    "frameImages": [{
      "inputImage": "c64351d5-4c59-42f7-95e1-eace013eddab",
      "frame": "first"
    }]
  },
  "positivePrompt": "The character begins to walk forward with natural movement",
  "resolution": "720p",
  "duration": 5,
  "settings": {
    "audio": false
  }
}
{
  "taskType": "videoInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da8",
  "model": "prunaai:p-video@0",
  "positivePrompt": "A person speaking to camera with natural gestures and expressions",
  "inputs": {
    "audio": "b4c57832-2075-492b-bf89-9b5e3ac02503"
  },
  "resolution": "1080p",
  "fps": 24,
  "settings": {
    "audio": true
  }
}
{
  "taskType": "videoInference",
  "taskUUID": "b8c4d952-7f27-4a6e-bc9a-83f01d1c6d64",
  "model": "prunaai:p-video@0",
  "positivePrompt": "A dynamic product reveal with smooth camera movement",
  "width": 1280,
  "height": 720,
  "duration": 6,
  "seed": 42,
  "settings": {
    "draft": true,
    "promptUpsampling": false
  }
}
Ask AI
Context: Full page
Send feedback
Context: Full page