Chat, Conversations, and Images
Start conversations, send images, dictate text, and manage chats.
Follow the bold button names and numbered steps. Technical names are included only when you need to recognize a model or file inside the app.
Start and manage conversations
Goal
Create a clean conversation without losing older chats.
Steps
- Select New chat in the conversation sidebar.
- Type a message and press Enter or select Send.
- To rename a conversation, use its rename action, type the title, then press Enter.
- Press Escape to cancel a rename.
- To remove a conversation, choose its delete action and confirm Delete this chat?.
Expected result
Each conversation remains available for the active character until deleted. Removing the final conversation creates a new empty one.
If something goes wrong
Local Waifu avoids creating repeated empty chats. Send a message in the current empty conversation before selecting New chat again.
Write and send messages
By default:
- Enter sends;
- Shift+Enter inserts a new line;
- Settings → Appearance & Chat → Send on Enter changes this behavior;
- the send action is disabled for empty text and while a send is already in progress.
Select Stop to end the visible response stream. The local model may still finish the current generation internally, so CPU or GPU use can remain high briefly.
Use Command/Ctrl+F to search chat history.
Attach an image
Goal
Send one image for a vision-capable model to inspect.
Before you start
- Use a normal image file under 5 MB.
- Configure a local vision model or a cloud model that supports images.
- Only one pending image can be attached at a time.
Steps
- Select the image/attachment button or drag an image onto the composer.
- Review the preview.
- Add a question or instruction.
- Send the message.
Expected result
The image and text are sent to the configured vision route and appear in the conversation.
If something goes wrong
- That file isn’t an image: choose an image rather than a PDF or document.
- Image is too large: reduce it below 5 MB.
- Couldn’t read that image: export it again as PNG or JPEG.
- If vision is unavailable, install a vision-capable model or configure a compatible cloud provider.
Dictate without sending automatically
Dictation records speech, transcribes it, and inserts the text into the message field. It does not send the message automatically. Review and edit the transcription, then send it yourself.
A dictation session can run for up to ten minutes. If the microphone disconnects or the time limit is reached, keep the text already captured and start another session.