vision_ocr.py

from asyncio import run
from avalan.entities import GenerationSettings
from avalan.model.vision.image import ImageTextToTextModel
from os.path import isfile
from sys import argv, exit


async def example(path: str) -> None:
    print("Loading model... ", end="", flush=True)
    with ImageTextToTextModel("google/gemma-3-12b-it") as vm:
        print(f"DONE. Running image recognition for {path}", flush=True)

        text = await vm(
            path,
            "Transcribe the text on this image, word for word, "
            "keeping format when possible.",
            settings=GenerationSettings(max_new_tokens=1024),
        )

        print(text, flush=True)


if __name__ == "__main__":
    path = (
        argv[1]
        if len(argv) == 2 and isfile(argv[1])
        else exit(f"Usage: {argv[0]} <valid_file_path>")
    )
    run(example(path))