Musa Aria

A cloned voice that can laugh, sigh and gasp

Most text to speech reads the words and stops there. Musa Aria performs them. Type a sound tag where you want it, and your cloned voice laughs, sighs, gasps or clears its throat right there in the line.

Try it free Hear before and after

9sound tags

1language: English

3×faster than real time on a GPU

2.9 GBdownload

Before and after

Same voice, now with feeling

Before · the input

“Hi, I'm Sarah. I read stories out loud every night, and honestly, I think a good voice can make any little moment feel special.”

After · Musa Aria

“[chuckle] Okay, I honestly did not expect that to work on the first try.”

“[sigh] It has been a very long day, but the episode is finally finished.”

“Wait, you recorded all of this at home? [laugh] That is incredible.”

Musa Aria on a GPU. The tags are in the script exactly as shown. The input voices are AI-generated. Only clone voices you own or have permission to use.

Why Aria

What it does best

Sounds written into the script

Add a tag from the Sound menu, or type it. The sound lands exactly where you put it, in the same voice.

Quick on a GPU

On a mid-range NVIDIA card, Aria made 24.6 seconds of speech in 7.8 seconds. It also runs on the CPU, at about real-time speed on a fast desktop.

Clones from one clip

Give Aria five seconds or more of a clear voice. It uses the first few seconds to learn how that person sounds.

Saved voices are instant

Voices in My Voices are prepared once and kept, so they skip the set-up next time.

Sound tags

Write the sound into the script

Choose a tag from the Sound menu, or type it where it should happen.

[laugh][chuckle][sigh][gasp][cough][clear throat][sniff][groan][shush]

Controls

Shape the result

Temperature: Move from consistent to expressive
Sound menu: Insert any of the 9 tags at the cursor
Pause and say-as markers: Control pacing, dates, numbers and units
Voice sample: At least 5 seconds of clear speech. Aria listens to the first 15 seconds.
Where to find it: Voice Clone Pro

Hardware

What it needs

Download: 2.9 GB
Windows, CPU only: Works, slowly · 8 GB RAM, 16 GB recommended
Windows with NVIDIA: Best · 4 GB GPU memory
Apple Silicon Mac: Runs on the CPU · 16 GB recommended
Speed: Measured on an RTX 4060 Ti: 24.6 s of speech in 7.8 s. On that PC’s CPU: 22.9 s of speech in 24 s.

Aria is English only and needs a reference of at least 5 seconds. Audio from Aria carries an inaudible AI watermark.

FAQ

Musa Aria questions

Which sounds can Musa Aria make?

[laugh], [chuckle], [sigh], [gasp], [cough], [clear throat], [sniff], [groan] and [shush].

Do I need the GPU pack?

No, but it helps a lot on Windows. Aria also runs on the CPU. It just takes longer, about as long as the audio itself on a fast desktop and more on a laptop.

Does Aria work on a Mac?

Yes. On Apple Silicon it runs on the CPU, so it is slower than on an NVIDIA GPU. We recommend 16 GB of memory.

What is the audio watermark?

Aria adds an inaudible watermark that marks the audio as AI-generated. You can’t hear it, and it stays on.