NtTestAlert/tortoise-tts
1
0
Fork 0
forked from mrq/tortoise-tts
Code Issues Pull Requests Packages Projects Releases Wiki Activity

cleaned up element order with Blocks, also added preset updating the samples/iterations counts

Browse Source
This commit is contained in:
mrq 2023-02-05 03:53:46 +00:00
parent 078dc0c6e2
commit d2aeadd754
2 changed files with 121 additions and 79 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
README.md
View File

@ -8,6 +8,10 @@ Similar to my own findings for Stable Diffusion image generation, this rentry ma
I link those a bit later on as alternatives for Windows+AMD users. You're free to skip the installation section and jump after that. I link those a bit later on as alternatives for Windows+AMD users. You're free to skip the installation section and jump after that.
>\> Wheres the love for Linux abloobloo
I'm extremely lazy and can't be assed to install Arch Linux again, much less create shell script equivalents. The commands should be almost 1:1 with what's in the batch file, save for the line to activate the venv.
>\>Ugh... why bother when I can just abuse 11.AI? >\>Ugh... why bother when I can just abuse 11.AI?
I very much encourage (You) to use 11.AI while it's still viable to use. For the layman, it's easier to go through the hoops of coughing up the $5 or abusing the free trial over actually setting up a TorToiSe environment and dealing with its quirks. I very much encourage (You) to use 11.AI while it's still viable to use. For the layman, it's easier to go through the hoops of coughing up the $5 or abusing the free trial over actually setting up a TorToiSe environment and dealing with its quirks.
@ -46,6 +50,14 @@ If you've done everything right, you shouldn't have any errors.
To check for updates with the Web UI, simply enter `git pull` in the command prompt, while the TorToiSe workspace is the current working directory. To check for updates with the Web UI, simply enter `git pull` in the command prompt, while the TorToiSe workspace is the current working directory.
### Pitfalls You May Encounter
I'll try and make a list of "common" (or what I feel may be common that I experience) issues with getting TorToiSe set up:
* `failed reading zip archive: failed finding central directory`: You had a file fail to download completely during the model downloading initialization phase. Please open either `%USERPROFILE%\.cache\tortoise\models\` or `%USERPROFILE%\.cache\huggingface\models\`, and delete the offending file.
You can deduce what that file is by reading the stack trace. A few lines above the last like will be a line trying to read a model path.
* `torch.cuda.OutOfMemoryError: CUDA out of memory.`: You most likely have a GPU with low VRAM (~4GiB), and the small optimizations with keeping data on the GPU is enough to OOM. Please open the `start.bat` file and add `--low-vram` to the command (for example: `py app.py --low-vram`) to disable those small optimizations.
## Preparing Voice Samples ## Preparing Voice Samples
Now that the tough part is dealt with, it's time to prepare voice sample clips to use. Now that the tough part is dealt with, it's time to prepare voice sample clips to use.
@ -73,24 +85,26 @@ Now you're ready to generate clips. With the command prompt still open, simply e
If you're looking to access your copy of TorToiSe from outside your local network, pass `--share` into the command (for example, `python app.py --share`). You'll get a temporary gradio link to use. If you're looking to access your copy of TorToiSe from outside your local network, pass `--share` into the command (for example, `python app.py --share`). You'll get a temporary gradio link to use.
You'll be presented with a bunch of options, but do not be overwhelmed, as most of the defaults are sane, but below are a rough explanation on which input does what: You'll be presented with a bunch of options, but do not be overwhelmed, as most of the defaults are sane, but below are a rough explanation on which input does what:
* `Text`: text you want to be read. You wrap text in `[brackets]` for "prompt engineering", where it'll affect the output, but those words won't actually be read. * `Prompt`: text you want to be read. You wrap text in `[brackets]` for "prompt engineering", where it'll affect the output, but those words won't actually be read.
* `Emotion`: the "emotion" used for the delivery. This is a shortcut to starting with `[I am really ${emotion}],` in your text box. I assume the emotion is deduced during the CLVP pass. * `Emotion`: the "emotion" used for the delivery. This is a shortcut to utilizing "prompt engineering" by starting with `[I am really <emotion>,]` in your prompt. This is not a guarantee, however.
* `Voice`: the voice you want to clone. You can select `custom` if you want to use input from your microphone. * `Custom Emotion + Prompt`: a non-preset "emotion" used for the delivery. This is a shortcut to utilizing "prompt engineering" by starting with `[<emotion>]` in your prompt.
* `Microphone Source`: Not required, unless you use `custom`. * `Voice`: the voice you want to clone. You can select `microphone` if you want to use input from your microphone.
* `Preset`: shortcut values for sample count and iteration steps. Use `none` if you want to provide your own values. Better presets rresult in better quality at the cost of computation time. * `Microphone Source`: Use your own voice from a line-in source.
* `Seed`: initializes the PRNG initially to this value, use this if you want to reproduce a generated voice. Currently, I don't have a way to expose the seed used.
* `Candidates`: number of outputs to generate, starting from the best candidate. Depending on your iteration steps, generating the final sound files could be cheap, but they only offer alternatives to the samples generated to pull from (in other words, the later candidates perform worse), so don't be compelled to generate a ton of candidates. * `Candidates`: number of outputs to generate, starting from the best candidate. Depending on your iteration steps, generating the final sound files could be cheap, but they only offer alternatives to the samples generated to pull from (in other words, the later candidates perform worse), so don't be compelled to generate a ton of candidates.
* `Autoregressive samples`: analogous to samples in image generation. More samples = better resemblance / clone quality, at the cost of performance. * `Seed`: initializes the PRNG to this value. Use this if you want to reproduce a generated voice.
* `Diffusion iterations`: influences audio sound quality in the final output. More iterations = higher quality sound. This step is relatively cheap, so do not be discouraged from increasing this. * `Preset`: shortcut values for sample count and iteration steps. Clicking a preset will update its corresponding values. Higher presets result in better quality at the cost of computation time.
* `Temperature`: how much randomness to introduce to the generated samples. Lower values = better resemblance to the source samples, but some temperature is still required for great output. This value definitely requires playing around depending on the voice you use. * `Samples`: analogous to samples in image generation. More samples = better resemblance / clone quality, at the cost of performance. This strictly affects clone quality.
* `Iterations`: influences audio sound quality in the final output. More iterations = higher quality sound. This step is relatively cheap, so do not be discouraged from increasing this. This strictly affects quality in the actual sound.
* `Temperature`: how much randomness to introduce to the generated samples. Lower values = better resemblance to the source samples, but some temperature is still required for great output. This value is very inconsistent and entirely depends on the input voice.
* `Diffusion Sampler`: sampler method during the diffusion pass. Currently, only `P` and `DDIM` are added, but does not seem to offer any substantial differences in my short tests.
After you fill everything out, click `Submit`, and wait for your output in the output window. The sampled voice is also returned, but if you're using multiple files, it'll return the first file, rather than a combined file. After you fill everything out, click `Run`, and wait for your output in the output window. The sampled voice is also returned, but if you're using multiple files, it'll return the first file, rather than a combined file.
All outputs are saved under `./result/[voice name]/[timestamp]/` as `result.wav`, with the text used saved alongside it. There doesn't seem to be an inherent way to add a Download button in Gradio, so keep that folder in mind. All outputs are saved under `./result/[voice name]/[timestamp]/` as `result.wav`, and the settings in `input.txt`. There doesn't seem to be an inherent way to add a Download button in Gradio, so keep that folder in mind.
To save you from headaches, I strongly recommend playing around with shorter sentences first to find the right values for the voice you're using before generating longer sentences. To save you from headaches, I strongly recommend playing around with shorter sentences first to find the right values for the voice you're using before generating longer sentences.
As a quick optimization, I modified the script to where the `conditional_latents` are saved after loading voice samples. If there's voice samples that have a modification time newer than this cached file, it'll skip loading it and load the normal WAVs instead. As a quick optimization, I modified the script to where the `conditional_latents` are saved after loading voice samples, and subsequent uses will load that file directly (at the cost of not returning the `Sample voice` to the web UI). If there's voice samples that have a modification time newer than this cached file, it'll skip loading it and load the normal WAVs instead.
## Example(s) ## Example(s)

162
app.py
View File

@ -9,7 +9,7 @@ from datetime import datetime
from tortoise.api import TextToSpeech from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_audio, load_voice, load_voices from tortoise.utils.audio import load_audio, load_voice, load_voices
def inference(text, emotion, prompt, voice, mic_audio, preset, seed, candidates, num_autoregressive_samples, diffusion_iterations, temperature, diffusion_sampler, progress=gr.Progress()): def generate(text, emotion, prompt, voice, mic_audio, preset, seed, candidates, num_autoregressive_samples, diffusion_iterations, temperature, diffusion_sampler, progress=gr.Progress()):
if voice != "microphone": if voice != "microphone":
voices = [voice] voices = [voice]
else: else:
@ -41,13 +41,6 @@ def inference(text, emotion, prompt, voice, mic_audio, preset, seed, candidates,
start_time = time.time() start_time = time.time()
presets = {
'Ultra Fast': {'num_autoregressive_samples': 16, 'diffusion_iterations': 30, 'cond_free': False},
'Fast': {'num_autoregressive_samples': 96, 'diffusion_iterations': 80},
'Standard': {'num_autoregressive_samples': 256, 'diffusion_iterations': 200},
'High Quality': {'num_autoregressive_samples': 256, 'diffusion_iterations': 400},
'None': {'num_autoregressive_samples': num_autoregressive_samples, 'diffusion_iterations': diffusion_iterations},
}
settings = { settings = {
'temperature': temperature, 'length_penalty': 1.0, 'repetition_penalty': 2.0, 'temperature': temperature, 'length_penalty': 1.0, 'repetition_penalty': 2.0,
'top_p': .8, 'top_p': .8,
@ -61,7 +54,7 @@ def inference(text, emotion, prompt, voice, mic_audio, preset, seed, candidates,
'diffusion_sampler': diffusion_sampler, 'diffusion_sampler': diffusion_sampler,
'progress': progress, 'progress': progress,
} }
settings.update(presets[preset])
gen, additionals = tts.tts( text, **settings ) gen, additionals = tts.tts( text, **settings )
seed = additionals[0] seed = additionals[0]
@ -97,68 +90,103 @@ def inference(text, emotion, prompt, voice, mic_audio, preset, seed, candidates,
seed seed
) )
def update_presets(value):
PRESETS = {
'Ultra Fast': {'num_autoregressive_samples': 16, 'diffusion_iterations': 30, 'cond_free': False},
'Fast': {'num_autoregressive_samples': 96, 'diffusion_iterations': 80},
'Standard': {'num_autoregressive_samples': 256, 'diffusion_iterations': 200},
'High Quality': {'num_autoregressive_samples': 256, 'diffusion_iterations': 400},
}
if value in PRESETS:
preset = PRESETS[value]
return (gr.update(value=preset['num_autoregressive_samples']), gr.update(value=preset['diffusion_iterations']))
else:
return (gr.update(), gr.update())
def main(): def main():
text = gr.Textbox(lines=4, label="Prompt") with gr.Blocks() as demo:
emotion = gr.Radio( with gr.Row():
["None", "Happy", "Sad", "Angry", "Disgusted", "Arrogant", "Custom"], with gr.Column():
value="None", text = gr.Textbox(lines=4, label="Prompt")
label="Emotion",
type="value",
)
prompt = gr.Textbox(lines=1, label="Custom Emotion + Prompt (if selected)")
preset = gr.Radio(
["Ultra Fast", "Fast", "Standard", "High Quality", "None"],
value="None",
label="Preset",
type="value",
)
candidates = gr.Slider(value=1, minimum=1, maximum=6, label="Candidates")
num_autoregressive_samples = gr.Slider(value=128, minimum=0, maximum=512, step=1, label="Samples")
diffusion_iterations = gr.Slider(value=128, minimum=0, maximum=512, step=1, label="Iterations")
temperature = gr.Slider(value=0.2, minimum=0, maximum=1, step=0.1, label="Temperature")
diffusion_sampler = gr.Radio(
["P", "DDIM"],
value="P",
label="Diffusion Samplers",
type="value",
)
voice = gr.Dropdown( emotion = gr.Radio(
os.listdir(os.path.join("tortoise", "voices")) + ["random", "microphone", "disabled"], ["None", "Happy", "Sad", "Angry", "Disgusted", "Arrogant", "Custom"],
label="Voice", value="None",
type="value", label="Emotion",
) type="value",
mic_audio = gr.Audio( interactive=True
label="Microphone Source", )
source="microphone", prompt = gr.Textbox(lines=1, label="Custom Emotion + Prompt (if selected)")
type="filepath", voice = gr.Dropdown(
) os.listdir(os.path.join("tortoise", "voices")) + ["microphone"],
seed = gr.Number(value=0, precision=0, label="Seed") label="Voice",
type="value",
)
mic_audio = gr.Audio(
label="Microphone Source",
source="microphone",
type="filepath",
)
selected_voice = gr.Audio(label="Source Sample") candidates = gr.Slider(value=1, minimum=1, maximum=6, label="Candidates")
output_audio = gr.Audio(label="Output") seed = gr.Number(value=0, precision=0, label="Seed")
usedSeed = gr.Textbox(label="Seed", placeholder="0", interactive=False)
interface = gr.Interface( preset = gr.Radio(
fn=inference, ["Ultra Fast", "Fast", "Standard", "High Quality", "None"],
inputs=[ value="None",
text, label="Preset",
emotion, type="value",
prompt, )
voice, num_autoregressive_samples = gr.Slider(value=128, minimum=0, maximum=512, step=1, label="Samples", interactive=True)
mic_audio, diffusion_iterations = gr.Slider(value=128, minimum=0, maximum=512, step=1, label="Iterations", interactive=True)
preset,
seed, temperature = gr.Slider(value=0.2, minimum=0, maximum=1, step=0.1, label="Temperature")
candidates, diffusion_sampler = gr.Radio(
num_autoregressive_samples, ["P", "DDIM"],
diffusion_iterations, value="P",
temperature, label="Diffusion Samplers",
diffusion_sampler type="value",
], )
outputs=[selected_voice, output_audio, usedSeed],
allow_flagging='never' prompt.change(fn=lambda value: gr.update(value="Custom"),
) inputs=prompt,
interface.queue().launch(share=args.share) outputs=emotion
)
preset.change(fn=update_presets,
inputs=preset,
outputs=[
num_autoregressive_samples,
diffusion_iterations,
],
)
with gr.Column():
selected_voice = gr.Audio(label="Source Sample")
output_audio = gr.Audio(label="Output")
usedSeed = gr.Textbox(label="Seed", placeholder="0", interactive=False)
submit = gr.Button(label="Generate")
submit.click(generate,
inputs=[
text,
emotion,
prompt,
voice,
mic_audio,
preset,
seed,
candidates,
num_autoregressive_samples,
diffusion_iterations,
temperature,
diffusion_sampler
],
outputs=[selected_voice, output_audio, usedSeed],
)
demo.queue().launch(share=args.share)
if __name__ == "__main__": if __name__ == "__main__":