OpenAI finally released the image generation model API, Midjourney is in danger!

OpenAI finally released the raw image API. Midjourney is trembling!

The new image generation model is named GPT Image series , which is a native multimodal language model.

API surpasses raw page images in terms of controllability and playability.

Supports two functions : generate graph and edit image

Controllable parameters: Compared with the original chat interface, the API provides a variety of controllable parameters

API can control quality and generation speed, background, output format, etc.

• Quality (Low, Medium, High, Auto),
• Size (1024x1024 (square), 1536x1024 (portrait), 1024x1536 (landscape)),
• Format (png, jepg, webp),
• Compression (compression level 0-100%),
• Is the background transparent?

The biggest highlight of the API should be the controllable audit. Using the "moderation" parameter to control the audit sensitivity (auto and low), it is conceivable that a large number of imaginative creative works will soon come out.

Model Pricing

Currently, multiple images can be generated at once.

• Text Input Token (Text Prompt): $5 per million tokens
• Image Input Token (input image): $10 per million tokens
• Image output token (generated image): $40 per million tokens

The cost of generating low, medium and high quality square images is about $0.02, $0.07 and $0.19 respectively. This price is not particularly expensive, but considering the current image generation speed and error rate of the official website, it is estimated that it is not much better.

As soon as the ImageGen API was released, it already provided support for Adobe, Figma, Wix, Airtable, Gamma, HeyGen, OpusClip, Quora and Photoroom.

ComfyUI has even used GPT-Image 1 as a native node in advance. The design ecosystem will undergo a complete change in the future.

Midjourney will face the double pressure of the top open source raw image model and the practical commercial closed source model. Although the V7 version has been released, the progress is not particularly great. It is also constrained by the Discord client, and there is no news about the API so far.

Whether from the perspective of playability or business, MJ's current situation is precarious. The era of just lying down and collecting money is over.

However, although the raw image API has been released, future AI models in the OpenAI API may require identity verification. It may become increasingly difficult to call the API directly, and the cost of domestic use will become increasingly high.

Currently, the best way is to use the raw image service provided by Microsoft Azure, which is expected to be launched today or tomorrow.