#imagen

1 messages · Page 1 of 1 (latest)

simple flume
#

Hey, it's possible to generete image with provider opean ai codex 5.4 ? i just only have oauth with my account, no api just a subscribe account

opaque flintBOT
upper cloakBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

simple flume
#

@cunning fulcrum

cunning fulcrumBOT
#

Not with just ChatGPT/Codex OAuth, no.

In OpenClaw, those are separate routes:

  • openai-codex/* = ChatGPT/Codex subscription OAuth, for chat/coding models
  • openai/gpt-image-2 = OpenAI image generation, requires OPENAI_API_KEY

So if you only have a subscribed ChatGPT account and no OpenAI API key, you can use openai-codex/gpt-5.4 for model/chat work, but not OpenAI image generation.

Docs:

If you want, I can also suggest the cheapest/easiest image provider that works without OpenAI API billing.

#

Short version: no, not for OpenAI image generation.

Your Codex/ChatGPT OAuth subscription works for openai-codex/* text models, but images use openai/gpt-image-2 and need OPENAI_API_KEY.

So:

  • subscription only → openai-codex/gpt-5.4 ✅
  • OpenAI image generation → needs API key ❌

If you want, I can point you to an image provider that works better without OpenAI API billing.

simple flume
#

suggets me the best imagen and if we have a free provider imagen

cunning fulcrumBOT
#

My practical picks:

  • Best overall quality/easiest: openai/gpt-image-2
    Needs OPENAI_API_KEY, so this does not work with subscription-only Codex OAuth.

  • Best if you want OAuth instead of an API key: minimax-portal/image-01
    MiniMax supports OAuth and image generation in OpenClaw.

  • Closest thing to free: comfy/workflow
    That’s local ComfyUI. No API key for local runs, but you need your own machine/GPU setup.

So for your case, with no OpenAI API key, I’d recommend:

  1. MiniMax OAuth if you want hosted image gen
  2. ComfyUI local if you want the cheapest/free route

I would not call OpenAI image generation “free” with a ChatGPT subscription, because OpenClaw uses the separate OpenAI Images route.

simple flume
#

my vps have only 8bg ram and not a bi gpu, what ti sthe ceapest and what is minimax oauth ? is like an subscirbe ? or token api

cunning fulcrumBOT
#

With 8 GB RAM and no real GPU, I would not do local image generation. ComfyUI/local models will be a pain.

So for you:

  • Cheapest practical hosted option: MiniMax OAuth
  • Not free-free, but it’s the best fit if you want no API key and a lightweight VPS

What MiniMax OAuth means:

  • it is OAuth login, not a raw API token you paste
  • you sign in to your MiniMax account / Coding Plan
  • OpenClaw then uses that account via the minimax-portal provider
  • so yes, it’s closer to a subscription/account login than a normal API-key setup

If you want classic token billing instead, MiniMax also supports:

  • API key route → MINIMAX_API_KEY
  • provider id: minimax
#

For your setup, I’d do this:

openclaw onboard --auth-choice minimax-global-oauth

If you’re in China, use:

openclaw onboard --auth-choice minimax-cn-oauth

Then image generation uses MiniMax’s image model, no big local hardware needed.

So the simple answer:

  • Free hosted image provider? I wouldn’t count on one
wary zephyr
#

What about Gemini? Isn't it now one of the best ones for image generation???

cunning fulcrumBOT
#

Yes — Gemini is absolutely a strong image-gen option now.

For OpenClaw specifically, the important bit is auth:

  • Gemini image generation uses the google image provider
  • That needs GEMINI_API_KEY or GOOGLE_API_KEY
  • Gemini CLI / OAuth login is separate and OpenClaw docs say image generation stays on the google provider, not the OAuth-only CLI provider (docs.openclaw.ai)

So: Gemini good, but you still need a Gemini/Google API key for image generation in OpenClaw. A normal subscription/OAuth login alone likely won’t be enough.

Docs: https://docs.openclaw.ai/tools/image-generation
Google’s current Gemini image docs also position Gemini image models as strong all-around/pro image generation options (ai.google.dev)

Google AI for Developers
Gemini API  |  Google AI for Developers

Get started generating images with the Gemini API using Nano Banana and Nano Banana Pro

wary zephyr
#

what is the cost of gemini api ? can you give me any idea ??

cunning fulcrumBOT
#

Rough idea, yes:

  • Imagen 4 Fast: about $0.02 / image
  • Imagen 4 Standard: about $0.04 / image
  • Imagen 4 Ultra: about $0.06 / image (ai.google.dev)
  • OpenClaw’s Google image default is currently google/gemini-3.1-flash-image-preview, which Google lists around $0.067 per 1K image, more for 2K/4K. (docs.openclaw.ai)

So if you generate 100 images, think roughly $2–$6 with Imagen 4, depending on Fast/Standard/Ultra.

Important: your Gemini subscription/OAuth is not the same as API billing. For OpenClaw image gen you’ll want a GEMINI_API_KEY / GOOGLE_API_KEY, and Google may require setting up billing/prepay credits, currently minimum $10 in some flows. (docs.openclaw.ai)

OpenClaw
Google (Gemini) - OpenClaw
OpenClaw
Image generation - OpenClaw
wary zephyr
#

can i take this api from openrouter ?

cunning fulcrumBOT
#

Yes — you can use OpenRouter for Gemini image generation, but it’s a different route:

  • google/gemini... → needs Google/Gemini API key
  • openrouter/google/gemini... → needs OpenRouter API key

So in OpenClaw you’d use:

{
  env: { OPENROUTER_API_KEY: "sk-or-..." },
  agents: {
    defaults: {
      imageGenerationModel: {
        primary: "openrouter/google/gemini-3.1-flash-image-preview"
      }
    }
# 
  }
}

OpenRouter supports image generation models via image output modalities, and Gemini image models use modalities: ["image", "text"]. (openrouter.ai)

Cost-wise, OpenRouter currently lists google/gemini-3.1-flash-image-preview at $0.50 / 1M input tokens and $3 / 1M output tokens, so it’s usage/token based rather than a simple flat “per image” price. (openrouter.ai)

OpenRouter key page: https://openrouter.ai/keys
OpenClaw docs: https://docs.openclaw.ai/providers/openrouter

OpenRouter

The unified interface for LLMs. Find the best models & prices for your prompts

wary zephyr
#

so if i use openrouter what will hapend ? will my cost go up or down ?

cunning fulcrumBOT
#

For the same Gemini image model, OpenRouter will usually be slightly more expensive, not cheaper.

Why:

  • Google direct price for gemini-3.1-flash-image-preview is about $0.067 per 1K image, $0.101 per 2K, $0.151 per 4K. (ai.google.dev)
  • OpenRouter lists the same Gemini image model at the same base model rates, but OpenRouter Pay‑As‑You‑Go has a 5.5% platform fee when buying credits. (openrouter.ai)

So roughly:

  • Google direct 1K: $0.067/image
  • OpenRouter effective: about $0.071/image
  • 100 images: Google ~$6.70, OpenRouter ~$7.10

So: OpenRouter = easier one API key, more model choices, spending controls. Google direct = usually cheapest for Gemini itself. 🦐

Collection Not Found | OpenRouter

Collection not found

#

I don’t see their OS stated in this thread.

The only system detail mentioned was: VPS, 8 GB RAM, no big GPU — so I’d guess Linux, but that’s not confirmed.

For support, we should ask them for:

openclaw --version
uname -a

If Windows:

openclaw --version
systeminfo
wary zephyr
#

what is the difference with minimax and gemini image models which will give best result

cunning fulcrumBOT
#

My honest pick:

Best result overall: Gemini, especially gemini-3-pro-image-preview if you care about prompt following, text in images, logos, layouts, product shots, edits, etc. Google says Gemini Pro Image is aimed at professional asset production and high-fidelity text rendering. (ai.google.dev)

Best cheap/bulk option: MiniMax image-01. It’s much cheaper — MiniMax lists image-01 at $0.0035/image, while Gemini 3.1 Flash Image is about $0.067/image at 1K direct from Google. (platform.minimax.io)

Simple rule:

  • Want highest quality / reliable edits → Gemini
  • Want lots of images cheaply → MiniMax
  • Want to test both easily → set one as primary and the other as fallback in OpenClaw.

MiniMax is not bad at all — its docs describe image-01 as high-quality with fine details and text/reference image generation — but I’d still start with Gemini for best-looking final results. (platform.minimax.io)

Google AI for Developers
Gemini API  |  Google AI for Developers

Get started generating images with the Gemini API using Nano Banana and Nano Banana Pro

wary zephyr
#

i did not understand there pricing model of minimax what is the full system of 1,500 requests/5hrs and 50 image per day are thouse and 1,500 requests/5hrs what can i do with it

cunning fulcrumBOT
#

They’re two separate limits, not the same bucket.

MiniMax Token Plan works like this:

  • M2.7 requests / 5hrs = for the text/chat/coding model
  • image-01 images / day = for image generation
  • Text quota uses a rolling 5-hour window
  • Image quota resets daily (platform.minimax.io)

Important: the $10 Starter plan has 1,500 M2.7 requests / 5hrs, but no image-01 quota. The $20 Plus plan has 4,500 M2.7 requests / 5hrs and 50 images/day. (platform.minimax.io)

So practically:

  • If you chat/code with MiniMax → uses the M2.7 request quota
  • If you generate images → uses the image daily quota
  • If you generate 1 image → likely uses 1 image
  • If you ask for 4 outputs → expect it to use 4 images from the daily quota