OpenAI Adds Transparent Background Generation to GPT-Image-2 API
OpenAI is previewing a transparent background feature for GPT-Image-2 through its API, letting developers generate PNGs with no background baked in at generation time. The company claims this produces cleaner results than traditional background removal, particularly on difficult edges like glass or thin fibers.
OpenAI is previewing transparent background support for GPT-Image-2 through its API, allowing developers to generate PNG images with no background rather than removing it afterward.
The feature is activated with a new background=transparent parameter in API calls. According to OpenAI, generating the alpha channel directly during image creation produces better results than conventional background-removal techniques applied after the fact. The company specifically cites improved handling of difficult areas such as transparent glass and thin fibers — regions where standard cutout tools typically struggle with edge artifacts.
OpenAI's accompanying Cookbook documentation outlines four target use cases:
- Product shots for e-commerce listings
- Diagrams for presentation software like PowerPoint
- Design elements including icons and stickers
- Merchandise artwork
Implementation details
OpenAI recommends against including background descriptions in prompts when using the transparent parameter, noting the model may still render a background if one is described in the text prompt despite the transparency setting.
For use cases involving charts or diagrams with specific numerical values, OpenAI advises manual verification of generated figures. The company states that GPT-Image-2 produces raster graphics and does not guarantee pixel-perfect accuracy for text or numbers embedded in images.
Developers looking to implement the feature need Python, the OpenAI and Pillow libraries, and an active API key. No pricing changes were mentioned alongside this update, and OpenAI has not disclosed whether transparent-background generation carries different token or compute costs than standard GPT-Image-2 requests.
What this means
This is a targeted workflow improvement rather than a new model or capability tier. Transparent background generation addresses a persistent pain point in AI image generation pipelines: post-hoc background removal tools frequently introduce artifacts around fine details like hair, glass, or fabric fibers, requiring manual cleanup before assets are usable in production contexts like e-commerce or marketing.
By baking transparency into the generation step itself, OpenAI is competing more directly with established design-asset workflows that currently rely on a generate-then-cut-out pipeline using separate tools. This matters most for teams building automated content pipelines — product photography, marketing collateral, or app icon generation — where eliminating a manual editing step reduces both cost and turnaround time.
The caveat about numerical accuracy in charts is a reminder that GPT-Image-2 remains a diffusion-based raster model, not a vector or data-rendering tool. Teams generating infographics or data visualizations still need a verification step before shipping generated assets, transparent background or not.
Related Articles
OpenAI tells developers to rewrite system prompts for GPT-6.2 Astra and Sol as defaults shift
OpenAI says developers moving to GPT-6.2 Astra or Sol should review their system prompts. The company published copy-ready templates in the models' documentation, and the Responses API now discards earlier reasoning content if conversation history is edited.
OpenAI releases GPT-Live-1.1 in the API, with standby mode billing idle WebRTC time at $0.02 per minute
OpenAI has released GPT-Live-1.1 in the API. The full-duplex voice model improves multilingual accents and transcription, and a new standby mode bills idle WebRTC time at $0.02 per minute instead of $0.05.
GPT-6.2 Astra scores 41.2% on XDOF's Ember-0 robot benchmark, up from 19.6% for GPT-6 Astra
Robotics infrastructure company XDOF reports that GPT-6.2 Astra reached a 41.2% average success rate across the 16 tasks of its Ember-0 benchmark, versus 19.6% for GPT-6 Astra. It is the first model to complete rope knot-tying and T-shirt unfolding on the benchmark. Average prediction latency fell from 15 seconds to 10.1 seconds.
OpenAI says GPT-6.2 Astra safeguards cut reasoning-extraction attack success to 0.2%
OpenAI announced on November 20, 2026 stronger defenses against adversarial distillation. It reports that adaptive attacks extracting protected reasoning from an earlier request succeed 0.2% of the time on GPT-6.2 Astra with updated safeguards, versus 54.0% on GPT-6 Astra with original safeguards.
Comments
Loading...