Qwen-Image-Edit on Novita AI: Professional-Grade Editing for Just $0.02 Per

Novita AI is proud to announce the launch of its Qwen-Image-Edit service, delivering professional-grade image editing powered by the 20-billion-parameter Qwen-Image model—now available for only $0.02 per image.

By combining semantic control (Qwen2.5-VL) and appearance control (VAE encoder), Qwen-Image-Edit makes it possible to perform precise, flexible, and efficient edits. From IP transformations and style changes, to localized text editing in English and Chinese, to fine-grained appearance adjustments—Novita AI brings the full power of state-of-the-art image editing into your workflow at an affordable cost.

What is Qwen-Image-Edit？

Qwen-Image-Edit Architecture

Qwen-Image-Edit is the image-editing variant of the 20-billion-parameter Qwen-Image model. It extends Qwen-Image’s advanced text rendering capabilities into editing tasks. It adopts a dual-path input design: routing the source image into both Qwen2.5-VL (for semantic control) and a VAE encoder (for appearance control), enabling precise and flexible editing.

1. Qwen2.5-VL Path (Semantic Control)

• What it is: Qwen2.5-VL is a multimodal vision-language model within the Qwen series. It specializes in understanding text prompts and the overall semantics of images.
• What it enables: High-level semantic control—such as changing styles, replacing objects, or rotating viewpoints—while ensuring semantic consistency across edits.

2. VAE Encoder Path (Appearance Control)

• What it is: A Variational Autoencoder (VAE) is a common image encoder used in generative models. It compresses the input image into a latent representation.
• What it enables:
  • Preserves low-level details of the original image (color, texture, local shapes).
  • Ensures that untouched regions remain fully consistent during local edits, avoiding “spillover” effects or unintended changes in unrelated areas.

What is Qwen-Image-Edit‘s Functionality?

1. Semantic Editing

Enables major transformations such as IP conversion, object rotation (including novel 90°/180° view synthesis), and style changes—all while preserving semantic consistency.

2. Appearance Editing

Supports adding, removing, or modifying visual elements (e.g., adding signs with reflections, deleting stray hairs, changing clothing or backgrounds) while keeping untouched regions fully intact.

3. Precise Text Editing

Allows bilingual (Chinese & English) text insertion, deletion, or modification in images, while preserving font, size, and overall visual style—ideal for localized poster or headline edits.

Qwen-Image-Edit Benchmark

What are the System Requirements for Qwen-Image-Edit?

Qwen launchs a DFloat11 lossless compression of the original Qwen/Qwen-Image-Edit BF16 weights. It reduces model size by ~32% while producing bit-identical outputs and enabling efficient GPU inference. With DFloat11, Qwen-Image-Edit runs on a single 32 GB GPU, or on a single 24 GB GPU with CPU offloading, without any loss in quality.

Run Qwen-Image-Edit on your own Novita AI GPU instance at ultra-low cost and start building your AI app today:

• RTX 5090 (32 GB VRAM) — 16 vCPU, 96 GB RAM — $0.50/hr
• L40S (48 GB VRAM) — 28 vCPU, 125 GB RAM — $0.55/hr
• A100 SXM (80 GB VRAM) — 14 vCPU, 240 GB RAM — $1.60/hr
• H100 SXM (80 GB VRAM) — 16 vCPU, 128 GB RAM — $1.80/hr

One-click Deploy, scale up to 8 GPUs per instance, and keep full control over your environment—perfect for fast prototyping or production workloads.

Start to Deploy your Instance

Compare Qwen-Image-Edit and Stable Diffusion, Nano Banana, DALL·E 4, Photoshop

Feature / Tool	Qwen-Image-Edit	Stable Diffusion	Nano Banana	DALL·E 4	Photoshop
Ease of Use	Plug‑and‑play with text prompts for editing	Flexible but needs prompt tuning	Very easy to use in google ecosystem	Need pro subscription	Steep learning curve; manual tools
Editing Style	Precise semantic & appearance editing; excellent text handling	Great for generation/inpainting	Integrate various elements	Fantastic for ideation, advertising concepts, and art creation.	Manual control; reliable but manually intensive
Speed	Slower generation; depends on hardware	Slower generation; depends on hardware	Very fast	About 1 Minutes	Very fast for manual workflows
Text Editing Capabilities	Excellent—including bilingual, English and Chinese	Poor; especially weak with Chinese or complex layouts	Not Mention	Not very accurate	Excellent (if fonts/elements available)

1. If you need accurate text editing (signs, posters, bilingual content)

• ✅ Qwen-Image-Edit → Best choice. Handles English + Chinese text precisely, preserves fonts/styles, and edits text seamlessly.
• ❌ Stable Diffusion / DALL·E 4 → Struggle with accurate text.
• ✅ Photoshop → Works if you already have fonts/elements and don’t mind manual editing.

2. If you prioritize speed & convenience

• ✅ Nano Banana → Lightning-fast and very easy to use inside Google’s ecosystem. Great for quick iterations, character consistency, and consumer workflows.
• ✅ Photoshop → Instant manual edits (if you’re skilled).
• ❌ Qwen-Image-Edit / Stable Diffusion → Slower, hardware-dependent generation.
• ❌ DALL·E 4 → Around 1 minute per image, not suitable if you need rapid turnaround.

3. If you want creativity, ideation, and concept art

• ✅ DALL·E 4 → Fantastic for advertising concepts, art style exploration, and ideation.
• ✅ Stable Diffusion → Flexible for inpainting & style mixing if you’re willing to tune prompts or fine-tune models.
• ❌ Qwen-Image-Edit → Better at precise edits than freeform creativity.
• ❌ Photoshop → Creative but manual; slower for ideation at scale.

4. If you need precise local edits & professional control

• ✅ Qwen-Image-Edit → Excellent for semantic edits + appearance preservation, e.g., swapping clothes, removing details, rotating objects.
• ✅ Photoshop → Gold standard for pixel-level manual control.
• ❌ Nano Banana / DALL·E 4 → Less suited for fine-grained local control.

5. If ease of use matters most

• ✅ Nano Banana → Simplest, embedded in Google ecosystem, low friction.
• ✅ Qwen-Image-Edit → Prompt-based, plug-and-play.
• ❌ Stable Diffusion → Requires model management & prompt expertise.
• ❌ Photoshop → Steep learning curve, manual effort.

Is Qwen-Image-Edit Suitable for Professional Use?

Qwen-Image-Edit delivers studio-quality results that make it highly suitable for professional photography, marketing materials, and commercial design projects.

Benchmark performance: Published results show state-of-the-art accuracy across multiple editing benchmarks, validating its consistency and reliability in demanding workflows.

Fine-grained editing: As VentureBeat notes, it “gives Photoshop a run for its money”, excelling at detailed tasks like posters, signs, T-shirts, and calligraphy where text precision really matters.

Text editing excellence: Unlike many generative models, it supports bilingual (Chinese + English) precise text edits, maintaining font, size, and style—critical for localized commercial content like advertising campaigns or branded assets.

Best Practices for Qwen-Image-Edit

Novita launches the Qwen-Image-Edit API, with pricing at just $0.02 per image.

Step 1: Log In and Access the Model Library

Log in to your account and click on the Model Library button.

Step 2: Choose Your Model

Browse through the available options and select the model that suits your needs.

Step 3: Get Your API Key

To authenticate with the API, we will provide you with a new API key. Entering the “Settings“ page, you can copy the API key as indicated in the image.

Step 4: Install the API

Install API using the package manager specific to your programming language.

After installation, import the necessary libraries into your development environment. Initialize the API with your API key to start interacting with Novita AI LLM. This is an example of using chat completions API for python users.

Vidu 2.0 Image to Video API Example

import requests

url = "https://api.novita.ai/v3/async/qwen-image-edit"

payload = {
    "prompt": "<string>",
    "image": "<string>",
    "seed": 123,
    "output_format": "<string>"
}
headers = {
    "Content-Type": "<content-type>",
    "Authorization": "<authorization>"
}

response = requests.post(url, json=payload, headers=headers)

print(response.json())

Extract Image URL

import requests

url = "https://api.novita.ai/v3/async/task-result"

headers = {
    "Content-Type": "<content-type>",
    "Authorization": "<authorization>"
}

response = requests.get(url, headers=headers)

print(response.json())

With the release of Qwen-Image-Edit, Novita AI has lowered the barrier for professional image editing. For just $0.02 per image, creators, developers, and businesses can now access studio-quality editing capabilities that rival traditional tools like Photoshop—while offering unique advantages in automation, bilingual text handling, and semantic precision.

Try Qwen-Image-Edit today and unlock the future of intelligent image editing.

Frequently Asked Questions

Novita AI is the All-in-one cloud platform that empowers your AI ambitions. Integrated APIs, serverless, GPU Instance — the cost-effective tools you need. Eliminate infrastructure, start free, and make your AI vision a reality.

Recommend Reading

Qwen Image Edit VS Nano Banana: Detailed User or Hands-Free

PixVerse V4.5 T2V on Novita AI: The Cheapest Way to Build Cinematic AI Videos

Unleashing the Power of Wan 2.2 I2V on Consumer Hardware