Pictures and voices

Rally works mostly in words, but it can also make two other things: pictures and recordings of a voice reading text aloud. Both come from its Multimodal generation plugin, which loads by itself when you ask. You do not need to switch anything on.

Drawing a picture#

Describe what you want and Rally draws it. While it works you see a row that says Generating an image.

Some requests that work well:

• "Draw a friendly cartoon owl reading a spreadsheet, for the top of our finance newsletter."
• "A flat, simple illustration of a rocket leaving a laptop screen, in our brand colours, navy and orange. Landscape."
• "Make a square icon of a coffee cup with a lightning bolt, on a white background."

The more you say, the closer the first attempt lands. Mention the style (photo, watercolour, flat illustration, line drawing), the colours, the mood, what should be in the picture, and what should not.

You can ask for a shape:

Pictures come out as PNG files.

Rally makes one picture per request. If you want three options to choose from, ask for three, and Rally makes them one after another. It is an enthusiastic artist rather than a fast one, so give it a moment.

Editing a picture you attach#

Attach an image (see Attaching files) and say what to change: "Make the sky a sunset", "Remove the coffee cup from the desk", "Turn this photo into a pencil sketch". The row says Editing an image.

If you only want part of the picture changed, you can attach a second image as a mask: a PNG that marks the area Rally should work on, leaving the rest alone. Say which attachment is the picture and which is the mask.

An edit does not make a new, separate file. It becomes a new version of the same file, so the original is still there in the file's history if you change your mind. Versions are explained in Artifacts.

Turning text into a voice recording#

Rally can read text aloud in a natural-sounding voice and save it as an MP3 recording. The row says Generating a voice recording.

Try:

• "Record this welcome message for our phone line, in a calm, warm voice."
• "Make an audio version of this week's update so I can listen on my commute."
• "Read this paragraph in an upbeat, energetic tone."

You can ask for a tone of voice (calm, cheerful, serious, slow and clear) and Rally passes that along. There are several voices to choose from, including alloy, ash, ballad, coral, echo, fable, nova, onyx, sage, shimmer, verse, marin and cedar. If you do not pick one, you get marin. The easiest way to choose is to ask for two short samples in different voices and pick the one you like.

Each request makes one recording, from up to about 4,000 characters of text (roughly a page and a half). For something longer, ask Rally to split it into parts.

Not the same as Speak#

This is different from the Speak button under a message, from Reading Rally's answers. Speak reads a message aloud to you, there and then, and leaves nothing behind. A voice recording is a file you keep, play again, download and share.

Where the results go#

Pictures and recordings are saved as files in the Channel, the same as any other file Rally makes, so everyone in the Channel can open them. To preview, keep one open beside the chat, or save it to your Mac, see Opening, docking and downloading.

What Rally cannot make#

• Video. Rally cannot make video clips, animations or moving pictures.
• More than one at a time. One picture or one recording per request, as above.

Next: Memory and coming back later.