Builder mode
Media
Choose who makes pictures, video and speech for your agent, and add the provider keys they need.
SettingsMedia controls who makes pictures, video and speech for your agent. In builder mode it's under "Models" in the Settings list, next to SettingsModel and SettingsTool keys.
Open SettingsMedia (open it). The card is headed "Pictures, speech and video".
What you can set
There are three rows:
| Row | What it controls |
|---|---|
| Image | The provider and model Hermes uses when it makes a picture |
| Video | The provider and model Hermes uses when it makes a video |
| Speech | The provider Hermes uses when it turns text into spoken audio |
Image and Video each have a "Provider" menu. Each row starts on "Your model: OpenAI", "Your model: xAI (Grok)" or the matching provider label when your model can make that kind of media. Otherwise it starts on "Off" until you save a choice. You also see providers with models for that row. If you pick a provider, a "Model" menu appears for that row.
Speech has one "Provider" menu. It's shared with SettingsVoice.
Your model
"Your model: OpenAI", for example, means Hermes can use the same account or key that powers your main model for that kind of media. The option appears only when your current model provider supports it.
| Current model provider | "Your model" can make |
|---|---|
| ChatGPT (Codex) sign-in | Image |
| Grok (SuperGrok) | Image and Video |
| xAI (API key) | Image and Video |
| OpenAI | Image |
| OpenRouter | Image and Video |
| DeepInfra | Image and Video |
When a row uses "Your model: OpenAI", for example, the row says "Uses your model's OpenAI key". With a ChatGPT or Grok sign-in it says "Uses your model's sign-in, no key needed". There's no separate key box in that row.
Providers
Image and Video use these provider choices:
| Provider | Image | Video | What it needs |
|---|---|---|---|
| Your model: OpenAI, for example | When listed | When listed | No extra key in Media |
| Off | Yes | Yes | Nothing |
| fal | Yes | Yes | fal key |
| OpenAI | Yes | No | OpenAI key |
| xAI (Grok) | Yes | Yes | xAI key |
| OpenRouter | Yes | Yes | OpenRouter key |
| DeepInfra | Yes | Yes | DeepInfra key |
| Krea | Yes | No | Krea key |
| Meta AI | Yes | No | Meta AI key |
ChatGPT (your sign-in) is used through "Your model: ChatGPT (your sign-in)" when your main model is ChatGPT (Codex). It's not listed as a separate provider choice.
Speech uses these provider choices:
| Provider | What it needs |
|---|---|
| OpenAI | OpenAI key |
| ElevenLabs | ElevenLabs key |
| Google Gemini | Google Gemini key |
| MiniMax | MiniMax key |
| Edge (free) | No API key needed |
Keys and billing
Each picture or video is billed to your own provider account. The Video row also warns that video can cost much more than pictures.
When an Image or Video provider needs a saved key and doesn't have one, the row says "Needs a key" and shows a password field. Speech shows a password field without that status line. Paste the key and press "Save media settings".
MyHermes checks saved keys for OpenAI, xAI, OpenRouter and ElevenLabs. If one of those providers clearly rejects the key, the form says that provider rejected it and keeps the old settings. If the check times out or gets any other answer, MyHermes still saves the key. For providers without a check here, MyHermes saves the key and the agent will find out when it tries to use it.
When a provider already has a saved key, the row shows "Remove key". Pressing it removes that saved key for the provider and you still need to press "Apply changes" in the dashboard banner.
OpenAI, xAI, ElevenLabs and any provider that also appears in SettingsVoice use the same saved key here and in Voice. If you remove one of those keys in Media, you remove it for Voice too. If you saved it in Voice, it shows as saved here.
Not available
Sometimes a provider choice is disabled and the form says "Not available while your model uses the same key setting (for example MyHermes AI)."
That means your agent has one setting for that key, and your model already uses it. The model key always wins that setting, so pick "Your model: OpenAI" if it's listed, or pick another provider.
Applying changes
After a save, the form says "Saved. Use the banner at the top to restart your agent and apply the changes." The dashboard banner says "Media settings changed. Apply to restart your agent with them." Press "Apply changes" in that banner.
If MyHermes restarts your agent, the banner changes to "Restarting your agent" and then "Your agent is back and ready." If your agent hasn't been set up yet, nothing restarts and the saved settings apply when your agent first starts.
Video, speech and chat
Pictures can show inline in the web chat. Video and audio don't play in the web chat yet. Hermes is set up to send a video or speech file to Telegram if Telegram is connected and say so in web chat. If Telegram isn't connected, it's set up to tell you where the file was saved and that the web chat can't play it yet.
Speech and Voice
The Speech row picks the text to speech provider. SettingsVoice uses the same provider.
SettingsVoice also has "Spoken replies". When it's on and you send Hermes a voice message on Telegram, it answers with a voice reply too. Web chat can't play those spoken replies yet. If no speech provider was saved yet, choosing one from Media starts spoken replies off. Turning the switch on or off is saved through the same Media settings path and also waits for the restart banner to apply.
What's next
- Models and keys: pick the model behind Hermes and manage your other provider keys.
- Chat: how pictures appear in web chat.
- Telegram: connect Telegram so Hermes can send media files there.