Text-to-Image
MiniMax H3
Diffusers
Safetensors
English
image-to-image
image-editing
image-generation
Instructions to use MiniMaxAI/MiniMax-H3-Image with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use MiniMaxAI/MiniMax-H3-Image with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("MiniMaxAI/MiniMax-H3-Image", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
File size: 2,788 Bytes
7b6142f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 | #!/usr/bin/env python3
"""Send reference1.png and prompt.edit.txt to a MiniMax-H3-Image server and write the PNG.
python3 edit.py [--url http://127.0.0.1:30000] [--out edit.png]
The images are sent in --image order, which defines <Picture 1>, <Picture 2>, ... in the
prompt. Uses only the Python standard library.
"""
import argparse
import base64
import json
import mimetypes
import urllib.request
import uuid
from pathlib import Path
# absolute(), not resolve(): in the Hugging Face cache this file is a symlink into blobs/.
HERE = Path(__file__).absolute().parent
def multipart(fields: dict, images: list) -> tuple:
boundary = f"minimax-h3-image-{uuid.uuid4().hex}"
parts = []
for name, value in fields.items():
parts.append(f'--{boundary}\r\nContent-Disposition: form-data; name="{name}"\r\n\r\n{value}\r\n'.encode())
for path in images:
content_type = mimetypes.guess_type(path.name)[0] or "application/octet-stream"
parts.append(
f'--{boundary}\r\nContent-Disposition: form-data; name="image"; filename="{path.name}"\r\n'
f"Content-Type: {content_type}\r\n\r\n".encode()
)
parts.append(path.read_bytes() + b"\r\n")
parts.append(f"--{boundary}--\r\n".encode())
return b"".join(parts), f"multipart/form-data; boundary={boundary}"
def main() -> None:
parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument("--url", default="http://127.0.0.1:30000", help="server base URL")
parser.add_argument("--image", action="append", type=Path,
help="reference image; repeat for <Picture 2>, ... (default: reference1.png)")
parser.add_argument("--prompt", type=Path, default=HERE / "prompt.edit.txt", help="prompt file")
parser.add_argument("--size", default="4096x4096", help="WIDTHxHEIGHT")
parser.add_argument("--seed", type=int, default=1234)
parser.add_argument("--out", default="edit.png")
args = parser.parse_args()
fields = {
"prompt": args.prompt.read_text(),
"size": args.size,
"seed": str(args.seed),
"generator_device": "cpu",
"response_format": "b64_json",
"output_format": "png",
}
body, content_type = multipart(fields, args.image or [HERE / "reference1.png"])
request = urllib.request.Request(
f"{args.url.rstrip('/')}/v1/images/edits", data=body, headers={"Content-Type": content_type}
)
with urllib.request.urlopen(request, timeout=1800) as response:
payload = json.load(response)
with open(args.out, "wb") as image:
image.write(base64.b64decode(payload["data"][0]["b64_json"]))
print(f"wrote {args.out}")
if __name__ == "__main__":
main()
|