roberthein/modelname-new - AI Language Models Tool

Overview

modelname-new is a public image-generation model hosted on Replicate that exposes a full API schema for text-to-image, image-to-image, and inpainting workflows. The model accepts a text prompt plus optional image and mask inputs, supports adjustable denoising steps, guidance scale, prompt strength for img2img, and custom width/height or preset aspect ratios. According to the model's API page, it also supports loading external LoRA weights (HuggingFace, CivitAI, or arbitrary .safetensors URLs) and provides a fast quantized inference mode (go_fast) for lower-latency runs. ([replicate.com](https://replicate.com/roberthein/modelname-new/versions/4f30f0d7ea1adeee4f1d8bb5dc4015655fef5cf181659fe5c545f988845ee788/api)) The Replicate model page reports this entry is open and runnable via the site or the API, with approximately 233 recorded runs and a noted inference environment using Nvidia H100 hardware. Replicate lists an estimated cost of about $0.029 per prediction on their hosted infrastructure and typical prediction latency around 19 seconds, and the model can also be run locally via Docker since it is open-source. The model lacks a README on the public page and the visible version metadata identifies the latest version by its SHA-like id. ([replicate.com](https://replicate.com/roberthein/modelname-new?utm_source=openai))

Key Features

• Text-to-image generation with adjustable guidance and inference steps.
• Image-to-image (img2img) with prompt_strength for controlled edits.
• Mask-based inpainting using an uploaded mask image.
• Supports loading external LoRA weights (HuggingFace, CivitAI, or .safetensors).
• Fast quantized 'go_fast' mode (fp8) for lower-latency inference.

Example Usage

Example (python):

import replicate

# Replace with your Replicate API token in environment or pass token directly
# This example runs the specific published version and waits for synchronous output.
model_ref = "roberthein/modelname-new:4f30f0d7ea1adeee4f1d8bb5dc4015655fef5cf181659fe5c545f988845ee788"

client = replicate.Client()

output = client.run(
    model_ref,
    input={
        "prompt": "A photorealistic portrait of a red fox wearing a leather jacket, cinematic lighting",
        "aspect_ratio": "1:1",
        "num_inference_steps": 28,
        "guidance_scale": 3.0,
        "num_outputs": 1,
        "output_format": "webp",
        # Optional: load a LoRA weight (HuggingFace or CivitAI URL or <owner>/<model>)
        # "replicate_weights": "fofr/flux-pixar-cars",
        # "go_fast": True  # enable fp8 quantized fast path
    },
)

print("Generated output URIs:", output)

# Notes:
# - The model supports img2img by passing an `image` URL and a `prompt_strength` value.
# - You can also call the HTTP API directly via POST to /v1/predictions using the version id.
# See the model API schema on Replicate for all available input fields. ([replicate.com](https://replicate.com/roberthein/modelname-new/versions/4f30f0d7ea1adeee4f1d8bb5dc4015655fef5cf181659fe5c545f988845ee788))

Pricing

Replicate shows an approximate hosted cost of $0.029 per prediction for this model; cost may vary by input size and settings. You can also run the open-source model locally via Docker. ([replicate.com](https://replicate.com/roberthein/modelname-new?utm_source=openai))

Benchmarks

Recorded runs on Replicate: 233 runs (Source: https://replicate.com/roberthein/modelname-new)

Estimated cost per run (Replicate hosted): $0.029 per run (approx.) (Source: https://replicate.com/roberthein/modelname-new)

Typical prediction latency: ≈19 seconds (Source: https://replicate.com/roberthein/modelname-new)

Hosted hardware: NVIDIA H100 (reported) (Source: https://replicate.com/roberthein/modelname-new)

Input / capability highlights: Supports prompt, img, mask, LoRA weights, custom aspect_ratio, go_fast mode (Source: https://replicate.com/roberthein/modelname-new/versions/4f30f0d7ea1adeee4f1d8bb5dc4015655fef5cf181659fe5c545f988845ee788/api)

Key Information

• Category: Language Models
• Type: AI Language Models Tool