GPT-6.1 Sol is not available from our provider yet. Prompt refinement currently uses GPT-6 Sol.
Describe the image you want. Sol shapes the instructions; GPT Image 2.5 renders the result.
Two models handle different parts of the job. You can inspect the instruction before spending credits on an image.
Currently, GPT-6 Sol interprets a rough idea and turns it into a concrete visual brief: subject, framing, light, materials, style and mood.
The current GPT-6 Sol connection writes text, not image pixels.
GPT Image 2.5 Flare renders the instruction into a visual result. The site's image editor can also use a reference photo for a later change.
If Sol is unavailable, your original idea still reaches the image model.
Each example records a plain idea, GPT-6.1 Sol’s rewritten instruction, and the image rendered by GPT Image 2.5. Try the same idea to create a fresh result.
GPT Image 2.5 outputPlain idea: “A plain skincare box in soft daylight, no logo or lettering”
A single unbranded skincare box stands upright on a pale stone surface, centered in a clean composition with ample negative space. Soft daylight enters from the left, casting a faint shadow. The box has a matte ivory finish with crisp folded edges and no logo, lettering, or graphics. Minimalist product photography with a calm, understated mood, shot straight on with a 50mm lens and shallow depth of field.
Recorded GPT-6 Sol rewrite · GPT Image 2.5 Flare render · Low quality
GPT Image 2.5 outputPlain idea: “A garden station above a canyon at sunset”
A small garden station perched on the rim of a vast canyon, with a sheltered platform, benches, and planted beds overlooking the drop. Frame the station in the foreground and the canyon stretching into the distance beneath a low horizon. Warm sunset light grazes the stone paths and foliage, casting long shadows. Weathered timber, pale sandstone, and green vines contrast with terracotta cliffs and a peach-colored sky. Atmospheric, tranquil landscape illustration with a wide-angle view.
Recorded GPT-6 Sol rewrite · GPT Image 2.5 Flare render · Low quality
GPT Image 2.5 outputPlain idea: “An editorial poster about urban cycling with the exact headline “RIDE LIGHT””
An editorial poster about urban cycling, featuring a cyclist riding through a city street, framed vertically with the cyclist in the lower half and the exact headline “RIDE LIGHT” large across the upper half. Low morning sunlight casts long shadows and catches the edges of buildings. Asphalt gray, warm concrete, and muted amber form a restrained palette. Bold editorial typography and clean geometric shapes create an energetic, contemporary mood. Slightly low-angle, wide-lens perspective.
Recorded GPT-6 Sol rewrite · GPT Image 2.5 Flare render · Low qualityStart naturally. Name the subject, setting and any exact words that need to appear.
Review the rewritten prompt. Copy it, edit your idea, or choose your original wording.
Select shape and quality, then create. Your image task and result stay in My Library.
Follow the recorded handoff below: a short idea becomes a GPT-6 Sol instruction, then GPT Image 2.5 renders the result.
“A plain skincare box in soft daylight, no logo or lettering.”
A single unbranded skincare box stands upright on a pale stone surface, centered in a clean composition with ample negative space. Soft daylight enters from the left, casting a faint shadow. The box has a matte ivory finish with crisp folded edges and no logo, lettering, or graphics. Minimalist product photography with a calm, understated mood, shot straight on with a 50mm lens and shallow depth of field.
Recorded GPT-6 Sol rewrite used for the image shown here.
No. Sol shapes text instructions, while GPT Image 2.5 renders the image. GPT-6.1 Sol is not available from our provider yet; this workflow currently uses GPT-6 Sol for prompt refinement.
Sol interprets your idea and writes a visual instruction. GPT Image 2.5 takes that instruction and renders the pixels.
Prompt refinement does not deduct credits. Image generation follows the selected quality rate. A guest may get one square, Low-quality preview shared with the site’s guest image trial, subject to daily capacity. Google sign-in gives new accounts 12 starter credits.
Describe your idea, review the instruction Sol shapes, then render it with GPT Image 2.5. You can keep your original wording instead.
No. ImgKick runs this workflow with its own server-side model connections. A ChatGPT subscription is not needed.
Signed-in users can choose square, landscape or portrait and Low, Auto, Medium, High, Xhigh or Max quality. The current image prices are 4, 10, 10, 20, 35 and 70 credits respectively. A guest preview is square and Low only.
Yes. Use the ImgKick image editor to upload a photo and describe the change.
Browse the ChatGPT Image 2.5 prompt library for tested prompts and real ImgKick image previews.