Can my Hermes make images, and who pays for them?
Yes, once you choose an image provider. Hermes Agent has an image tool, image_generate, but it shows that tool to the model only when an image provider is ready. Until then, Hermes acts as though it can’t make pictures. You choose the provider, and that provider bills you directly. Through OpenRouter, a per-image model such as FLUX 3 Klein costs $0.009 for a 1K image. Through FAL, Hermes’s default FLUX 2 Klein 9B costs $0.006 per megapixel. A ChatGPT or Codex sign-in uses your ChatGPT plan instead of an API bill. A new Vavio computer has no image provider set up.
This follows Hermes Agent v0.23.3, commit caafe53, checked on 22 November 2026. We reviewed official documentation and source, including how Vavio sets up Hermes, and read prices from FAL and OpenRouter that day. We did not generate any images, so this article makes no claim about image quality or speed.
Why does my Hermes say it can’t make images?
Usually because no image provider is ready. The image tool is switched on by default. Hermes leaves it out of the model’s tool list until one of these is true:
- A FAL key is set (
FAL_KEY), and you haven’t chosen another provider. - You are signed in to Nous Portal with access to its Tool Gateway, and you haven’t chosen another provider. A paid subscription includes the gateway. Some accounts also have a small free tool allowance.
- You chose another provider in Hermes’s tool settings, and its key or sign-in works.
A key you already have for chat doesn’t switch images on by itself. This applies to an OpenRouter, OpenAI or xAI key alike. Hermes checks only the provider you selected, so that a key you added for chat never moves you onto a paid image service you didn’t pick. If the provider you chose has no working key or sign-in, the tool is normally hidden. Hermes doesn’t quietly switch to another provider.
On a Vavio computer the setup writes no image settings and no image keys. It also turns off the free guest Nous identity that ordinary Hermes creates on first use. A guest identity doesn’t qualify for the gateway’s image tools anyway. So a new computer can’t make images until you choose a provider.
Which provider should I use, and what does it cost?
| Provider in Hermes | What you need | Who bills you | Price we read on 22 Nov 2026 |
|---|---|---|---|
| OpenRouter | Your OpenRouter key | OpenRouter credits | Per model: FLUX 3 Klein $0.009 per 1K image; Seedream 5.0 Flash $0.018 per image |
| FAL.ai | A FAL key | FAL | FLUX 2 Klein 9B (Hermes’s default) $0.006 per megapixel; other models up to $0.25 an image in Hermes’s list |
| Nous Subscription | Nous Portal sign-in with Tool Gateway access | Your Nous subscription | Billed per use against the subscription. The free allowance covers the FAL models only |
| OpenAI (Codex auth) | ChatGPT or Codex sign-in | Your ChatGPT plan | No API bill. Hermes doesn’t say how this counts toward your plan’s limits |
| OpenAI | An OpenAI API key | OpenAI API usage | By tokens. This is separate from a ChatGPT subscription |
| xAI (Grok) | An xAI key or Grok sign-in | xAI | Not checked |
These are the main options. Hermes v0.23.3 also has Krea, DeepInfra and Meta backends. The prices change, so check them with each provider before relying on them. FAL charges by image size: a 1024×1024 picture is about one megapixel, so Hermes’s default model costs well under a cent. Among the fixed prices in Hermes’s FAL list, the top is Recraft V4 Pro at $0.25 an image; the newer GPT Image models there are billed by tokens. On FAL, Hermes always asks for GPT Image at medium quality, which is about $0.03–0.06 a picture.
On OpenRouter, pick a model. If you leave the model unset, Hermes uses OpenAI’s GPT-5.4 Image 2 and falls back to Gemini 3 Pro Image. OpenRouter bills both by tokens, at $30 and $120 per million image-output tokens. Hermes chose them for quality, not price. If price matters, choose a per-image model such as FLUX 3 Klein in the model list. Its price depends on size: $0.008 at 768 pixels up to $0.036 at 4K.
The ChatGPT/Codex sign-in route has a catch. That backend treats the requested quality and size as suggestions. A portrait request can come back square, and Hermes reports what it actually received. For exact control over quality and size, use an API key route instead.
Your chat model is billed separately from all of this. The chat turns around a picture cost what they normally cost on your model; the image fee comes on top.
How do I set it up?
- In the Hermes app, open Capabilities and find image generation under See, hear and create. Pick a provider, paste its key if it needs one, and choose a model from the list.
- From a terminal,
hermes toolsopens the same choice under Image Generation. - Then ask in plain words, for example “make a square image of a lighthouse at dawn”. You give it a description and a shape (landscape, square or portrait), and optionally a picture to work from. It has no model setting of its own, so every picture uses the model you chose.
A key is a secret, so enter it in Hermes’s own settings, not in a chat message. See where to put a secret your Hermes needs.
Some models can also edit a picture you give them, such as a photo or a product shot. Pass the image along with your request. Models that only create new pictures refuse it with a clear error. To give Hermes a picture to work from, see what Hermes can read from a file you send.
Where does the picture end up?
With most providers Hermes saves the picture on the computer under ~/.hermes/cache/generated/images/. Hermes doesn’t clear that folder on the 24-hour schedule it uses for other temporary media. FAL models, used directly or through Nous, are the exception: they return a link to a copy FAL hosts, and Hermes’s documentation warns that such links expire after hours or days. If you want to keep one of those pictures, ask Hermes to download it into a folder. To get any file onto your own device, see how to get a file Hermes made.
Checking what you spent: Hermes records token-billed image calls in its usage figures (hermes insights and the dashboard’s Usage analytics). That covers token-billed models on OpenRouter and the OpenAI API. Per-picture models, including FLUX 3 Klein on OpenRouter, and FAL, xAI and Krea report nothing back to Hermes, so check the provider’s own dashboard. With Nous, the Nous Portal dashboard shows usage by tool.
What changes on a Vavio computer?
Vavio doesn’t add an image provider or pay for pictures. A new computer starts with none, and any provider you choose bills your own account. If your Hermes already uses OpenRouter for chat, the simplest start is to choose OpenRouter for images too: the key is already there, so you only pick a model. The Capabilities screen is in the same Hermes app that Vavio opens for you, so you can set this up without a terminal. The picture is made and saved on a computer that stays online, so a long batch carries on after you close your laptop.
Vavio provides a managed Hermes computer for $14.99 per month, with model and image usage billed separately by your providers. Compare the current hosting offer, or read what Hermes costs to run each month.
Sources
- Hermes v0.23.3 image generation guide, checked 22 November 2026
- How Hermes decides the image tool is available, checked 22 November 2026
- Hermes v0.23.3 Nous Tool Gateway guide, checked 22 November 2026
- Hermes's OpenRouter image backend and its default models, checked 22 November 2026
- Hermes's ChatGPT/Codex sign-in image backend, checked 22 November 2026
- FAL price for FLUX 2 Klein 9B, checked 22 November 2026
- OpenRouter image models and prices, checked 22 November 2026
- Vavio hosting offer, checked 22 November 2026
Written by Vavio with AI assistance and checked against pinned official Hermes v0.23.3 documentation and source, against Vavio's own computer setup source, and against provider price lists read on 22 November 2026. This is a source review. We did not generate images or judge image quality.