GPT Image 2: AI Image Editor Online
Bottom line
GPT Image 2 editing: up to 10 reference images, same 1K/2K/4K pricing as the generator (3/5/8 credits). Open the Image Edit tool, select GPT Image 2, upload your image, and the exact credit cost appears before the edit runs.
GPT Image 2 I2I is the editing half of OpenAI's GPT Image 2 — same underlying model, same resolution-based pricing as the text-to-image side, applied to an existing image instead of a blank prompt. You can attach up to 10 reference images per job, choose from 11 aspect ratios including auto, and render at 1K, 2K, or 4K for 3, 5, or 8 credits respectively. As on the generator, leaving aspect ratio on auto forces 1K, and a 1:1 output can't go to 4K.
Because it shares OpenAI's GPT multimodal comprehension, it handles long, multi-part editing instructions well — "replace the background, keep the subject's pose exactly, add the logo top-right" tends to come back as three distinct, correctly applied changes rather than one blended guess. Prompts run up to 20,000 characters.
It carries VdoBloom's STRICT content label, reflecting OpenAI's conservative provider filters, and holds a 4.6 user rating. GPT Image 2 (both directions) launched 2026-04-21, and the credit cost for your chosen resolution is shown before you generate.
What it costs on VdoBloom
Resolution-based, matching the generator: 3 credits at 1K, 5 at 2K, 8 at 4K, for any aspect ratio except 1:1 at 4K (unsupported — caps at 2K). A drafting workflow of two 1K attempts (6 credits) followed by a 2K final (5 credits) runs 11 credits total, comfortably inside a small credit top-up.
GPT Image 2 I2I vs GPT Image 1.5 I2I
GPT Image 1.5 is simpler and flat-priced: 10 credits regardless of resolution, 3 aspect ratios, 3,000-character prompts. GPT Image 2 trades that simplicity for range: 11 aspect ratios including auto, a 20,000-character prompt budget, and per-resolution pricing that undercuts 1.5 at 1K and 2K but costs more at 4K. Pick 2 for format flexibility and long instructions; pick 1.5 if your workflow already standardizes on its narrower ratio set.
Common editing problems and how to avoid them
The two most common surprises are resolution-related: leaving aspect ratio on auto silently caps output at 1K, and a 1:1 output can't be pushed to 4K no matter what resolution is selected. Neither produces an error message that explains itself, so if a result looks smaller than expected, check both settings before assuming the edit failed. For multi-part instructions, splitting the request into clearly separated clauses (rather than one long blended sentence) reduces the chance that one part of the edit gets dropped.
What creators use GPT Image 2 for
- Multi-step edits on an existing photo — background swap, object addition, text overlay — issued as one detailed instruction.
- Composites built from up to 10 reference images: a product shot, a background plate, and supporting style references in one job.
- Draft edits at 1K (3 credits) before committing to a 4K (8 credits) final pass.
How to use GPT Image 2 on VdoBloom
- 1Open the Image Edit tool and select GPT Image 2.
- 2Upload the image(s) to edit and describe the change you want.
- 3Review the credit cost and generate — results appear in your library.
Frequently asked questions
How much does GPT Image 2 I2I cost?
3 credits at 1K, 5 at 2K, 8 at 4K — the same resolution-based pricing as the text-to-image side of GPT Image 2.
How many reference images can I attach to one edit?
Up to 10 per job, which the model treats as combined context for a single instruction.
Is GPT Image 2 I2I free to try?
Your 10 free signup credits cover three 1K edits, or one 2K edit with 5 credits left over.
GPT Image 2 vs GPT Image 1.5 — which editor should I use?
GPT Image 1.5 charges a flat 10 credits regardless of resolution and offers only 3 aspect ratios (1:1/2:3/3:2). GPT Image 2 is resolution-tiered (3/5/8 credits) with 11 ratios including auto, and at 1K or 2K it's cheaper than 1.5's flat rate — only 4K costs more.
How strict is GPT Image 2 I2I's content filter?
STRICT — VdoBloom's tightest tier, reflecting OpenAI's conservative provider rules. It's more likely to decline sensitive or suggestive edit requests than models tagged HOT, such as Seedream 4.5 Edit.
What's the maximum resolution for editing?
4K, at 8 credits, with the same exception as the generator: a 1:1 square output can't render at 4K, so use 2K or a non-square ratio for the largest square-adjacent result.
My edit didn't apply, or GPT Image 2 isn't working — what should I check?
Confirm you're not combining a 1:1 aspect ratio with 4K (unsupported), and that aspect ratio is set explicitly if you need above 1K — auto forces 1K. If the job fails outright, VdoBloom refunds the credits automatically rather than charging for a non-result.
How do I prompt GPT Image 2 I2I for multi-step edits?
Break the instruction into explicit, separate changes — what to swap, what to keep unchanged, what to add — rather than one blended description. The 20,000-character prompt budget gives plenty of room to be literal.
GPT Image 2 vs Seedream 4.5 Edit — how do the prices compare?
At 1K or 2K, GPT Image 2 I2I (3 or 5 credits) is cheaper than Seedream 4.5 Edit's flat 7 credits. At 4K, GPT Image 2 costs more (8 credits) than Seedream 4.5's flat rate. Seedream 4.5 Edit also natively renders up to 4K without a separate resolution selection.