novita Python SDK

this SDK is based on the official API documentation

join our discord server for help

Installation

pip install novita-client

Examples

• fine tune example
• cleanup
• controlnet
• create-tile
• doodle
• img2img
• img2video
• instantid
• latent-consistency-txt2img
• lcm-img2img
• lcm-vs-txt2img
• make-photo
• merge-face
• mixpose
• model-search
• outpainting
• reimagine
• remove-background
• remove-text
• replace-background
• replace-object
• replace-sky
• txt2img-with-hiresfix
• txt2img-with-lora
• txt2img-with-refiner

Code Examples

cleanup

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.cleanup(
    image="https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png",
    mask="https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo_mask.png"
)

base64_to_image(res.image_file).save("./cleanup.png")

controlnet

#!/usr/bin/env python
# -*- coding: UTF-8 -*-

import os

from novita_client import NovitaClient, Img2ImgV3Request, Img2ImgV3ControlNetUnit, ControlnetUnit, Samplers, Img2ImgV3Embedding
from novita_client.utils import base64_to_image


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.img2img_v3(
    input_image="https://img.freepik.com/premium-photo/close-up-dogs-face-with-big-smile-generative-ai_900101-62851.jpg",
    model_name="dreamshaper_8_93211.safetensors",
    prompt="a cute dog, masterpiece, best quality",
    sampler_name=Samplers.DPMPP_M_KARRAS,
    width=512,
    height=512,
    steps=30,
    strength=1.0,
    controlnet_units=[
        Img2ImgV3ControlNetUnit(
            # image_base64="https://img.freepik.com/premium-photo/close-up-dogs-face-with-big-smile-generative-ai_900101-62851.jpg",
            image_base64="examples/fixtures/qrcode.png",
            model_name="control_v1p_sd15_brightness",
            preprocessor=None,
            strength=1
        )
    ],
    embeddings=[Img2ImgV3Embedding(model_name=_) for _ in [
        "BadDream_53202",
    ]],
    seed=-1,
)


base64_to_image(res.images_encoded[0]).save("./img2img-controlnet.png")

create-tile

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.create_tile(
    prompt="a cute flower",
)


base64_to_image(res.image_file).save("./create-tile.png")

doodle

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.doodle(
     image="https://img.freepik.com/premium-photo/close-up-dogs-face-with-big-smile-generative-ai_900101-62851.jpg",
     prompt="A cute dog",
)

base64_to_image(res.image_file).save("./doodle.png")

img2img

import pdb
import os

from novita_client import NovitaClient, Img2ImgV3ControlNetUnit, ControlNetPreprocessor, Img2ImgV3Embedding
from novita_client.utils import base64_to_image, input_image_to_pil
from concurrent.futures import ThreadPoolExecutor

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

res = client.img2img_v3(
    model_name="MeinaHentai_V5.safetensors",
    steps=30,
    height=512,
    width=512,
    input_image="https://img.freepik.com/premium-photo/close-up-dogs-face-with-big-smile-generative-ai_900101-62851.jpg",
    prompt="1 cute dog",
    strength=0.5,
    guidance_scale=7,
    embeddings=[Img2ImgV3Embedding(model_name=_) for _ in [
        "bad-image-v2-39000",
        "verybadimagenegative_v1.3_21434",
        "BadDream_53202",
        "badhandv4_16755",
        "easynegative_8955.safetensors"]],
    seed=-1,
    sampler_name="DPM++ 2M Karras",
    sd_vae="klF8Anime2VAE_klF8Anime2VAE_207314.safetensors",
    clip_skip=2,
    # controlnet_units=[Img2ImgV3ControlNetUnit(
    #     model_name="control_v11f1p_sd15_depth",
    #     preprocessor="depth",
    #     image_base64="./20240309-003206.jpeg",
    #     strength=1.0
    # )]
)

base64_to_image(res.images_encoded[0]).save("./img2img.png")

img2video

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URNOVITA_API_URII', None))
res = client.img2video(
    model_name="SVD-XT",
    steps=30,
    frames_num=25,
    image="https://replicate.delivery/pbxt/JvLi9smWKKDfQpylBYosqQRfPKZPntuAziesp0VuPjidq61n/rocket.png",
    enable_frame_interpolation=True
)


with open("test.mp4", "wb") as f:
    f.write(res.video_bytes[0])

instantid

import os
from novita_client import NovitaClient, InstantIDControlnetUnit
import base64



if __name__ == '__main__':
	client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

	res = client.instant_id(
		model_name="sdxlUnstableDiffusers_v8HEAVENSWRATH_133813.safetensors",
		face_images=[
			"https://raw.githubusercontent.com/InstantID/InstantID/main/examples/yann-lecun_resize.jpg",
		],
		prompt="Flat illustration, a Chinese a man, ancient style, wearing a red cloth, smile face, white skin, clean background, fireworks blooming, red lanterns",
		negative_prompt="(lowres, low quality, worst quality:1.2), (text:1.2), watermark, (frame:1.2), deformed, ugly, deformed eyes, blur, out of focus, blurry, deformed cat, deformed, photo, anthropomorphic cat, monochrome, photo, pet collar, gun, weapon, blue, 3d, drones, drone, buildings in background, green",
		id_strength=0.8,
		adapter_strength=0.8,
		steps=20,
		seed=42,
		width=1024,
		height=1024,
		controlnets=[
			InstantIDControlnetUnit(
				model_name='controlnet-openpose-sdxl-1.0',
				strength=0.4,
				preprocessor='openpose',
			),
			InstantIDControlnetUnit(
				model_name='controlnet-canny-sdxl-1.0',
				strength=0.3,
				preprocessor='canny',
			),
		],
		response_image_type='jpeg',
	)

	print('res:', res)

	if hasattr(res, 'images_encoded'):
		with open(f"instantid.png", "wb") as f:
			f.write(base64.b64decode(res.images_encoded[0]))

latent-consistency-txt2img

from novita_client import *
from novita_client.utils import save_image, read_image_to_base64, base64_to_image
import os
from PIL import Image
import random
import sys


def test_lcm_txt2img():
    client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

    x, y = 0, 0
    background = Image.new('RGB', (512 * 10, 512 * 10), (255, 255, 255))
    for i in range(20):
        animals = ["cat", "dog", "bird", "horse", "elephant", "giraffe", "zebra", "lion", "tiger", "bear", "sheep", "cow", "pig"]
        res = client.lcm_txt2img(
            prompt=f"a cute {random.choice(animals)}, masterpiece, best quality, realism",
            steps=8,
            image_num=5,
        )
        images = [base64_to_image(img.image_file) for img in res.images]
        for image in images:
            background.paste(image, (x, y))
            background.save("lcm.jpeg")
            x += 512
            if x >= 512 * 10:
                x = 0
                y += 512


def test_normal_txt2img():
    client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

    x, y = 0, 0
    background = Image.new('RGB', (512 * 10, 512 * 10), (255, 255, 255))
    for i in range(20):
        animals = ["cat", "dog", "bird", "horse", "elephant", "giraffe", "zebra", "lion", "tiger", "bear", "sheep", "cow", "pig"]
        res = client.sync_txt2img(
            Txt2ImgRequest(
                prompt=f"a cute {random.choice(animals)}, masterpiece, best quality, realism",
                steps=20,
                height=512,
                width=512,
                batch_size=5,
            )
        )
        images = [Image.open(BytesIO(b) for b in res.data.imgs_bytes)]
        for image in images:
            background.paste(image, (x, y))
            background.save("normal.jpeg")
            x += 512 
            if x >= 512 * 10:  
                x = 0
                y += 512


if __name__ == '__main__':
    if len(sys.argv) > 1:
        if sys.argv[1] == "normal":
            test_normal_txt2img()
        else:
            test_lcm_txt2img()

lcm-img2img

#!/usr/bin/env python
# -*- coding: UTF-8 -*-

import os
from novita_client import NovitaClient


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

res = client.lcm_img2img(
    model_name="dreamshaper_8_93211.safetensors",
    prompt="1 house",
    image="https://replicate.delivery/pbxt/JvLi9smWKKDfQpylBYosqQRfPKZPntuAziesp0VuPjidq61n/rocket.png",
    steps=4,
    guidance_scale=1,
    clip_skip=1,
    image_num=1,
)

print(res.to_json())
res.images[0]

lcm-vs-txt2img

from novita_client import *
from novita_client.utils import save_image, read_image_to_base64, base64_to_image
import os
from PIL import Image
import random
import sys


def test_lcm_txt2img():
    client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

    x, y = 0, 0
    background = Image.new('RGB', (512 * 10, 512 * 10), (255, 255, 255))
    animals = ["cat", "dog", "bird", "horse", "elephant", "giraffe", "zebra", "lion", "tiger", "bear", "sheep", "cow", "pig"]
    for i in range(20):
        object_prompt = animals[i % len(animals)]
        res = client.lcm_txt2img(
            prompt=f"a cute {object_prompt}, masterpiece, best quality, realism, high saturation",
            steps=8,
            image_num=5,
        )
        images = [base64_to_image(img.image_file) for img in res.images]
        for image in images:
            background.paste(image, (x, y))
            background.save("lcm.jpeg")
            x += 512
            if x >= 512 * 10:
                x = 0
                y += 512


def test_normal_txt2img():
    client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

    x, y = 0, 0
    background = Image.new('RGB', (512 * 10, 512 * 10), (255, 255, 255))
    animals = ["cat", "dog", "bird", "horse", "elephant", "giraffe", "zebra", "lion", "tiger", "bear", "sheep", "cow", "pig"]
    for i in range(20):
        object_prompt = animals[i % len(animals)]
        res = client.sync_txt2img(
            Txt2ImgRequest(
                model_name="dreamshaper_7_77036.safetensors",
                prompt=f"a cute {object_prompt}, masterpiece, best quality, realism",
                steps=20,
                height=512,
                width=512,
                batch_size=5,
            )
        )
        images = [Image.open(BytesIO(b)) for b in res.data.imgs_bytes]
        for image in images:
            background.paste(image, (x, y))
            background.save("normal.jpeg")
            # 更新位置计数器
            x += 512  # 向右移动一个图像的宽度
            if x >= 512 * 10:  # 如果一行已满，换到下一行
                x = 0
                y += 512


if __name__ == '__main__':
    if len(sys.argv) > 1:
        if sys.argv[1] == "normal":
            test_normal_txt2img()
        else:
            test_lcm_txt2img()

make-photo

#!/usr/bin/env python
# -*- coding: UTF-8 -*-

import os
from novita_client import NovitaClient, MakePhotoLoRA
import base64


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

res = client.make_photo(
    model_name="sd_xl_base_1.0.safetensors",
    prompt="anime artwork man img, portrait. anime style, key visual, vibrant, studio anime, highly detailed",
    negative_prompt="wrong, photo, deformed, black and white, realism, disfigured, low contrast",
    images=[
        "../testdataset/portrait2image/7.jpg"
    ],
    loras=[
        MakePhotoLoRA(
            model_name="sdxl_wrong_lora",
            strength=0.8
        )
    ],
    steps=25,
    guidance_scale=5,
    image_num=1,
    strength=0.3,
    seed=1024,
)


for idx in range(len(res.images_encoded)):
    with open(f"make_photo_{idx}.png", "wb") as f:
        f.write(base64.b64decode(res.images_encoded[idx]))

merge-face

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.merge_face(
    image="https://www.wgm8.com/wp-content/uploads/2016/06/images_wgm_online-only_Gaming_2016_30-06-16-1.jpg",
    face_image="https://p7.itc.cn/images01/20220220/285669b5682540a8a307a87d8745f530.jpeg",
)

base64_to_image(res.image_file).save("./merge_face.png")

mixpose

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.mixpose(
    image="https://image.uniqlo.com/UQ/ST3/my/imagesgoods/455359/item/mygoods_23_455359.jpg?width=494",
    pose_image="https://image.uniqlo.com/UQ/ST3/ca/imagesgoods/455359/item/cagoods_02_455359.jpg?width=494",
)

base64_to_image(res.image_file).save("./mixpose.png")

model-search

#!/usr/bin/env python
# -*- coding: UTF-8 -*-

from novita_client import NovitaClient, ModelType
# get your api key refer to https://docs.novita.ai/get-started/
client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

# filter by model type
print("lora count", len(client.models().filter_by_type(ModelType.LORA)))
print("checkpoint count", len(client.models().filter_by_type(ModelType.CHECKPOINT)))
print("textinversion count", len(
    client.models().filter_by_type(ModelType.TEXT_INVERSION)))
print("vae count", len(client.models().filter_by_type(ModelType.VAE)))
print("controlnet count", len(client.models().filter_by_type(ModelType.CONTROLNET)))


# filter by civitai tags
client.models().filter_by_civi_tags('anime')

# filter by nsfw
client.models().filter_by_nsfw(False)  # or True

# sort by civitai download
client.models().sort_by_civitai_download()

# chain filters
client.models().\
    filter_by_type(ModelType.CHECKPOINT).\
    filter_by_nsfw(False).\
    filter_by_civitai_tags('anime')

outpainting

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.outpainting(
    image="https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png",
    width=910,
    height=512,
    center_x=0,
    center_y=0,
)
base64_to_image(res.image_file).save("./outpainting.png")

reimagine

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.reimagine(
    image="/home/anyisalin/develop/novita-client-python/examples/doodle-generated.png"
)

base64_to_image(res.image_file).save("./reimagine.png")

remove-background

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.remove_background(
    image="https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png",
)
base64_to_image(res.image_file).save("./remove_background.png")

remove-text

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.remove_text(
    image="https://images.uiiiuiii.com/wp-content/uploads/2023/07/i-banner-20230714-1.jpg"
)

base64_to_image(res.image_file).save("./remove_text.png")

replace-background

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.replace_background(
    image="./telegram-cloud-photo-size-2-5408823814353177899-y.jpg",
    prompt="in living room, Christmas tree",
)
base64_to_image(res.image_file).save("./replace_background.png")

replace-object

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.replace_object(
    image="https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png",
    object_prompt="a dog",
    prompt="a cute cat"
)
base64_to_image(res.image_file).save("./replace_object.png")

replace-sky

import os

from novita_client import NovitaClient
from novita_client.utils import base64_to_image

client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.replace_sky(
    image="https://dynamic-media-cdn.tripadvisor.com/media/photo-o/17/16/a6/88/con-la-primavera-in-giappone.jpg?w=700",
    sky="galaxy"
)


base64_to_image(res.image_file).save("./replace_sky.png")

txt2img-with-hiresfix

import os

from novita_client import NovitaClient, Samplers, Txt2ImgV3HiresFix
from novita_client.utils import base64_to_image

from PIL import Image


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))
res = client.txt2img_v3(
    model_name='dreamshaper_8_93211.safetensors',
    prompt="a cute girl",
    width=384,
    height=512,
    image_num=1,
    guidance_scale=7.5,
    seed=12345,
    sampler_name=Samplers.EULER_A,
    hires_fix=Txt2ImgV3HiresFix(
        # upscaler='Latent'
        target_width=768,
        target_height=1024,
        strength=0.5
    )
)


base64_to_image(res.images_encoded[0]).save("./txt2img_with_hiresfix.png")

txt2img-with-lora

#!/usr/bin/env python
# -*- coding: UTF-8 -*-

import os
from novita_client import NovitaClient, Txt2ImgV3LoRA, Samplers, ProgressResponseStatusCode, ModelType, add_lora_to_prompt, save_image
from novita_client.utils import base64_to_image, input_image_to_pil
from PIL import Image


def make_image_grid(images, rows: int, cols: int, resize: int = None):
    """
    Prepares a single grid of images. Useful for visualization purposes.
    """
    assert len(images) == rows * cols

    if resize is not None:
        images = [img.resize((resize, resize)) for img in images]

    w, h = images[0].size
    grid = Image.new("RGB", size=(cols * w, rows * h))

    for i, img in enumerate(images):
        grid.paste(img, box=(i % cols * w, i // cols * h))
    return grid


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

res1 = client.txt2img_v3(
    prompt="a photo of handsome man, close up",
    image_num=1,
    guidance_scale=7.0,
    sampler_name=Samplers.DPMPP_M_KARRAS,
    model_name="dreamshaper_8_93211.safetensors",
    height=512,
    width=512,
    seed=1024,
)
res2 = client.txt2img_v3(
    prompt="a photo of handsome man, close up",
    image_num=1,
    guidance_scale=7.0,
    sampler_name=Samplers.DPMPP_M_KARRAS,
    model_name="dreamshaper_8_93211.safetensors",
    height=512,
    width=512,
    seed=1024,
    loras=[
        Txt2ImgV3LoRA(
           model_name="add_detail_44319",
           strength=0.9,
        )
    ]
)

make_image_grid([base64_to_image(res1.images_encoded[0]), base64_to_image(res2.images_encoded[0])], 1, 2, 512).save("./txt2img-lora-compare.png")

txt2img-with-refiner

import os

from novita_client import NovitaClient, Txt2ImgV3Refiner, Samplers
from novita_client.utils import base64_to_image
from PIL import Image


def make_image_grid(images, rows: int, cols: int, resize: int = None):
    """
    Prepares a single grid of images. Useful for visualization purposes.
    """
    assert len(images) == rows * cols

    if resize is not None:
        images = [img.resize((resize, resize)) for img in images]

    w, h = images[0].size
    grid = Image.new("RGB", size=(cols * w, rows * h))

    for i, img in enumerate(images):
        grid.paste(img, box=(i % cols * w, i // cols * h))
    return grid


client = NovitaClient(os.getenv('NOVITA_API_KEY'), os.getenv('NOVITA_API_URI', None))

r1 = client.txt2img_v3(
    model_name='sd_xl_base_1.0.safetensors',
    prompt='a astronaut riding a bike on the moon',
    width=1024,
    height=1024,
    image_num=1,
    guidance_scale=7.5,
    sampler_name=Samplers.EULER_A,
)

r2 = client.txt2img_v3(
    model_name='sd_xl_base_1.0.safetensors',
    prompt='a astronaut riding a bike on the moon',
    width=1024,
    height=1024,
    image_num=1,
    guidance_scale=7.5,
    sampler_name=Samplers.EULER_A,
    refiner=Txt2ImgV3Refiner(
        switch_at=0.7
    )
)

r3 = client.txt2img_v3(
    model_name='sd_xl_base_1.0.safetensors',
    prompt='a astronaut riding a bike on the moon',
    width=1024,
    height=1024,
    image_num=1,
    guidance_scale=7.5,
    sampler_name=Samplers.EULER_A,
    refiner=Txt2ImgV3Refiner(
        switch_at=0.5
    )
)


make_image_grid([base64_to_image(r1.images_encoded[0]), base64_to_image(r2.images_encoded[0]), base64_to_image(r3.images_encoded[0])], 1, 3, 1024).save("./txt2img-refiner-compare.png")