Spaces:

---

TTS-AGI

/

TTS-Arena

like 638

Running on CPU Upgrade

Hey,
Thanks for reaching out! We can definitely upgrade the model to v1.0. Has the PlayDialog-http model automatically been upgraded from Beta to v1.0 in the Python API?

Is it possible to relaunch it as a new model? Are you able to share the params you are currently passing through and the endpoint you are hitting?

If you're using our Python SDK (pyht), I recommend upgrading to the latest version and then calling tts() with voice_engine='PlayDialog' and protocol='http' instead of voice_engine='PlayDialog-http' as we changed the API to separate them. Otherwise it will be the same.

Is it possible to relaunch it as a new model? Are you able to share the params you are currently passing through and the endpoint you are hitting?

Yes, we can relaunch as a new model. Should it be labelled as PlayDialog 1.0?

The current params being used:

for chunk in play_client.tts(text, TTSOptions(voice="s3://voice-cloning-zero-shot/831bd330-85c6-4333-b2b4-10c476ea3491/original/manifest.json"), voice_engine="PlayDialog-http"):

That is the correct labeling! Will wait for @bryananderson to confirm -- but a question for you @mrfakename is how do you pick voices / do you cycle through them? Do you only pick one? Should we provide you a list?

Bumping @mrfakename

Can you share the params for the python sdk payload?

Hey @legofan94 ,
So sorry about the delay. Here's the code used for generation:

voice_engine = "PlayDialog"
tts_options = TTSOptions(voice="s3://voice-cloning-zero-shot/831bd330-85c6-4333-b2b4-10c476ea3491/original/manifest.json")
for chunk in play_client.tts(text, tts_options, voice_engine=voice_engine):
    if chunk == b'':
        play_client.close()
        break
    f.write(chunk)
    return f.name, None

Working on getting it added now! Are these the right params to use?

Almost! Can you use this voice instead of the one you have in params? It's a good neutral voice if you're only taking one. If you want to have different accents and speaker styles we can provide more.

@mrfakename

s3://voice-cloning-zero-shot/42c41808-0ddb-4674-8965-024a52ad6c8e/original/manifest.json

Makes sense, switched to that voice. Are there any other settings that need to be adjusted before launch?

Nope -- you're otherwise good!

Should be live shortly!