Images and voice
Generate images from a prompt, and dictate instead of typing.
Two features that use models other than the chat one, and bill the same way.
The image playground and microphone input both need Pro. Read out loud works on a free account. See Credits, Pro, and usage.
Images
Images is a playground for image models. It keeps its own list of pieces in the sidebar, the same shape as chats and notes, so earlier generations stay where you left them.
- Open Images and choose New Image.
- Describe what you want, or start from one of the suggestions.
- Pick the image model — the picker sits under the prompt, the way the chat model picker sits in the composer.
Image models are chosen separately from your chat model, so the model answering your questions and the model drawing your pictures need not be the same.
You can also generate an image without leaving a conversation: the Generate Image tool does it from chat, and can be switched off like any other tool.
Voice
Anywhere you can type a prompt — the chat composer and the image prompt — you can Dictate instead. Recording starts, and Stop & Transcribe turns it into text in the field, which you then send as normal.
Transcription uses a provider you choose under Settings → Default Models → Transcription Provider, and Settings → General holds the voice chat language.
What these cost
Both draw on the same balance as everything else: a request to a cloud model through 9xchat spends credits at that model's rate, a request with your own API key bills your provider, and a local model costs nothing.
Image generation is usually priced per image rather than per token, so it is worth checking the model's rate before a long session. See Credits, Pro, and usage.
Audio has no local option. Dictation sends your recording, and read out loud sends the text being spoken, to your audio provider — 123sudo or OpenAI. Both leave your machine even when your chat model is local. See Privacy and data flow.