Your brief: named speakers, their lines in order, and how the conversation moves
Direction for speaker names, turn order and pacing
What to leave out, such as overlapping turns or an extra speaker you never named
Vizify audio tools
AI audio tool
Give every speaker their lines and get the whole conversation as one track.
Give Vizify named speakers, their lines in order, and how the conversation moves. It routes to a model that exposes Dialogue mode, passes only the parameters that model supports, and returns one multi-speaker conversation in a single track.
Dialogue mode, checked before the job runs You direct speaker names, turn order and pacing Returns one multi-speaker dialogue audio artifact
Start with the messy versionOptional choices help Vizify shape the handoff.
Try a starting point
Your brief opens in Vizify for review before generation.
one multi-speaker dialogue audio artifact
one multi-speaker conversation in a single track
Ready to place in a podcast pilot, a training scenario, or a scripted product demo
What it does
Built to hand back finished work.
One route: Dialogue
This brief only reaches a model that exposes Dialogue mode, so the job asks for one multi-speaker conversation in a single track and nothing adjacent.
Dialogue parameters, inspected
Vizify reads the selected model's parameter list first and passes which speaker gets which voice, the turn order, and the pacing between turns only where the model supports them. Unsupported values are dropped, not approximated.
Tracked to the finished Dialogue file
Vizify follows the job to completion or failure and stores the audio, so you can check whether each speaker keeps the same voice across every turn before anyone else hears it.
FAQ
Questions, answered.
What does AI Dialogue Generator create?
One bounded job, one file: one multi-speaker conversation in a single track. It creates one dialogue track, not a complete podcast edit or perfectly consistent cloned voices.
Which model does Vizify use for Dialogue mode?
Vizify inspects a public model that exposes Dialogue mode and confirms it accepts which speaker gets which voice, the turn order, and the pacing between turns before sending anything. The choice happens at run time, so no model is promised in advance.
Is the audio ready immediately?
No. Audio generation is asynchronous: Vizify submits the request, tracks it while the model works, and reports completion or failure. The finished file is stored for playback.
Can I choose speaker names, turn order and pacing?
Ask for which speaker gets which voice, the turn order, and the pacing between turns in the brief. Each value is passed only when the inspected model lists support for it; the rest is left out rather than approximated.
Can I upload reference audio to shape the dialogue track?
Yes, with a free Vizify account: up to three files, 20 MB each, can steer the texture of the dialogue track. Use only audio you own.
Does Vizify clear the rights for the dialogue track?
No. Vizify clears no rights or licensing. You are responsible for the script, and each speaker gets a supported voice, not a real person's. Check the applicable terms before you publish or sell the result.
Do I need a Vizify account?
You can prepare the brief publicly; generation, billing, saving, and playback continue after sign-in.
What should I listen for before publishing the dialogue track?
Play it end to end and check whether each speaker keeps the same voice across every turn. The workflow does not promise per-speaker stems or a finished podcast edit, so plan an edit pass if you need that.
How this page works
You write named speakers, their lines in order, and how the conversation moves. Vizify turns that into one Dialogue-mode request, reads the selected model’s parameter list, and submits only the values it accepts. The audio is stored with the request, so one multi-speaker conversation in a single track is playable the moment the job succeeds.
What you control
The brief carries which speaker gets which voice, the turn order, and the pacing between turns. Write the exclusions too: naming overlapping turns or an extra speaker you never named up front is easier than fixing it later, since Vizify will not invent a parameter the model never exposed.
Before you publish
Play the result end to end and check whether each speaker keeps the same voice across every turn. It creates one dialogue track, not a complete podcast edit or perfectly consistent cloned voices. It does not promise per-speaker stems or a finished podcast edit, and it clears no rights. You are responsible for the script, and each speaker gets a supported voice, not a real person’s.
Brief to playable audio
Open Vizify