search

Plachtaa / seed-vc

State-of-the-Art zero-shot voice conversion & singing voice conversion with in context learning
GNU General Public License v3.0
663 stars 76 forks source link

Causes of Extended Audio Duration After Conversion and Solutions #40

Closed Bella-Tim closed 2 weeks ago

Bella-Tim commented 2 weeks ago

After conversion, the audio duration is different from the original, and it has become longer. What causes this issue? Besides adjusting the playback speed, are there any other methods to keep the converted audio duration the same as the original?

Plachtaa commented 2 weeks ago

Please share an example of before/after conversion, and how long does the conversion result is longer than original

Bella-Tim commented 2 weeks ago

For example, with this audio clip, I used a Trump model. The original was 4 seconds, but the converted version is 5 seconds. model audio: https://drive.google.com/file/d/1qHnQuMHSQ1RmlgasdoPAl7utqnN4H5Sp/view?usp=sharing original audio : https://drive.google.com/file/d/1O1AHT24W2_9bTQY_SWK5ighg5RHPmQF1/view?usp=sharing

Plachtaa commented 2 weeks ago

please share your converted audio as well, as the audio you shared here seems to only have original audio and reference audio

Bella-Tim commented 2 weeks ago

converted audio ： https://drive.google.com/file/d/1FtKZxO0KMKB8TJptKAig47GlGi3mZQh-/view?usp=sharing

Plachtaa commented 2 weeks ago

I am noticing your original audio has a length of 4.80 seconds, while the conversion result is of 5.80 seconds, and it sounds slower than source However, I tried conversion on the web app and get a result of 4.79 seconds. Please make sure you have correctly set the length adjust argument to 1.0

Bella-Tim commented 2 weeks ago

My bad, I overlooked this. I kept thinking I had set it to 1.0, but it was actually set to 1.2. Thank you.