Dialogue

Dialogue turns a script into one audio conversation, with a different voice for each character. You can redo any single line without making the whole thing again.

The page

The script fills the screen on the left, one line per turn. On the right, one panel has three tabs: Cast (voices and settings), Result and History. On a laptop the page itself doesn't scroll; the script and the panel do.

Make a conversation

1. Open Dialogue in the sidebar. A sample script is filled in.
2. Write your lines. Each line has:
   • the speaker, with a colour dot: choose the name to switch speaker, or + New speaker… to name a new one
   • what they say
   • an emotion and a volume, on the right of the line
3. Choose Add line under the script to add a line for the next speaker. Use the arrows to move a line up or down, and × to delete it.
4. In Cast, choose a voice for each speaker. The play button beside a voice plays a preview; choose the voice to search, filter and hear every voice.
5. Set Model, Language and Speed in Cast if you need to.
6. Choose Generate conversation. The price shows next to it.

The Result tab opens and plays the conversation when it's ready, with each line and its start time.

Paste a script instead

1. Choose Paste a script at the top right.
2. Paste lines like Host: Welcome to the show!, one line per turn.
3. Choose Use this script. It replaces the lines on the page.

How the script is read:

• Text before the first : becomes the speaker (up to 40 characters).
• A line without a speaker is added to the line before it.
• If the script starts without a speaker, that first line goes to "Narrator".

Fix one line

In Result, each line has two buttons:

• Redo line: say the same line again.
• Edit: change the text, emotion or volume, then choose Redo with changes.

Only that line is made again. The rest of the conversation stays as it was. The script keeps the new wording.

Options

• Emotion per line: Natural (the default), Neutral, Happy, Sad, Angry, Fearful, Disgusted or Surprised.
• Volume per line: 50%, 75%, 100%, 125%, 150% or 200%.
• Add a sound (under the script): puts a tag like (laughs) where your cursor last was, or at the end of the last line. Choices: Laughs, Sighs, Coughs, Clears throat, Gasps, Sniffs, Groans, Yawns.
• Model: Speech (1 credit per character) or Speech HD (2 credits per character).
• Language: Detect or one of 37 languages (the same list as Speech).
• Speed: 0.50× to 2.00×, for the whole conversation.
• Voices: your workspace's cloned and designed voices come first (marked yours), then Cinara's voices. Each new speaker gets a different voice to start with.
• Your workspace's pronunciations apply to every line.

Subtitles

• In Result, choose SRT or VTT.
• In History, choose SRT subtitles or VTT subtitles on any finished conversation.
• Subtitles come from the line timings and include speaker names.

History

The History tab lists your last 30 conversations, each with a player. Choose Open to load a conversation, its voices and settings back into the script and Result.

Limits

• Up to 200 lines.
• Up to 2,000 characters per line.
• Up to 50,000 characters in total.
• Speaker names up to 40 characters.
• Every line needs a speaker and some text before you can generate.

Credits

• Speech: 1 credit per character of all lines.
• Speech HD: 2 credits per character.
• Redo line and Redo with changes charge again for that one line's characters.
• If generation or a redo fails, those credits are refunded.

Tips

• Keep each line to one speaker's turn. Short turns sound more natural.
• Use emotions sparingly. Natural often works best.
• Use volume to make an aside quieter or a shout louder, instead of redoing the voice.
• Preview voices before you generate so each character sounds distinct.

Related

• Speech: one voice.
• Voice Library: find voices for your cast.
• Voice Cloning and Voice Design: make your own voices.
• Dubbing: translate an existing recording.