google colab about real-time-voice-cloning HOT 13 CLOSED

did somebody managed to get this working in Google Colab? Would really like to try but I don't have a gpu at home

I found this https://colab.research.google.com/github/ak9250/Real-Time-Voice-Cloning/blob/master/Real_Time_Voice_Cloning.ipynb

from real-time-voice-cloning.

did somebody managed to get this working in Google Colab? Would really like to try but I don't have a gpu at home

from real-time-voice-cloning.

Add this to line 1
%tensorflow_version 1.x

from real-time-voice-cloning.

Not from me, but I would certainly not mind linking it if someone makes one. I think a notebook-ish demo is doable

from real-time-voice-cloning.

@CorentinJ I see I have everything setup in colab but i think the problem is the demo_toolbox does not support non ui /running in cli yet?

I get this output running the demo_toolbox, demo_cli tests pass

WARNING: Logging before flag parsing goes to stderr.
W0701 03:24:00.135030 140347045406592 deprecation_wrapper.py:119] From /content/Real-Time-Voice-Cloning/synthesizer/models/modules.py:91: The name tf.nn.rnn_cell.RNNCell is deprecated. Please use tf.compat.v1.nn.rnn_cell.RNNCell instead.

Arguments:
datasets_root: /content/Real-Time-Voice-Cloning/UserAudio/speaker_01
enc_models_dir: encoder/saved_models
syn_models_dir: synthesizer/saved_models
voc_models_dir: vocoder/saved_models
low_mem: False

from real-time-voice-cloning.

No, the CLI demo is not finished yet. You will have to wait until then.

from real-time-voice-cloning.

@CorentinJ any update on this, once it works I can send a PR if you would like to add the colab notebook to the repo

from real-time-voice-cloning.

Yeah the CLI demo should be working, try it out.

from real-time-voice-cloning.

@CorentinJ just tried it got this error
/content/Real-Time-Voice-Cloning/UserAudio/speaker_01/audio_01.wav
Loaded file succesfully
Created the embedding
"this is a test"
Created the mel spectrogram
Synthesizing the waveform:
{| ████████████████ 57000/57600 | Batch Size: 6 | Gen Rate: 4.6kHz | }Caught exception: PortAudioError('Error querying device -1',)
Restarting

from real-time-voice-cloning.

@ak9250 This is because in demo_cli.py, when synthesizing completes, the synthesized audio will be played. However the Colab VM doesn't have any audio device. You need to modify the code a bit.

from real-time-voice-cloning.

@JasonWei512 yep that was it adding --no_sound as a argument fixes it thanks

from real-time-voice-cloning.

the tensor flow versions arent supporting in colab

from real-time-voice-cloning.

The colab notebook gives following error

Synthesizing the waveform:
{| ████████████████ 85500/86400 | Batch Size: 9 | Gen Rate: 11.7kHz | }generated_wav.dtype:  float64
Caught exception: AttributeError("module 'librosa' has no attribute 'output'")
Restarting

Fixed by - in the translate.py line 131 replace this

# librosa.output.write_wav(fpath, generated_wav.astype(np.float32), synthesizer.sample_rate)
import soundfile as sf
sf.write('output.wav', generated_wav.astype(np.float32), synthesizer.sample_rate, 'PCM_24')

from real-time-voice-cloning.