o
i @ sH d Z ddlZddlZddlZddlZddlZddlm Z ddl
m Z ej ddZ
e
jddd d
e
jd d d
g ddd e
jddd
g ddd e
jddeddd e
jdded d d e
Zed e ejejejejejd ZejZejZed d!Zd"ed#ed$ed%ed&ed'e f d(d)Z ej d*d+Z e d,e! d-e d. e" ej#d/d0b ej$d1d2d3d4d5d6Z%ej&d7d8d9d:Z'ej(d;d8d<1 ej)d=d>d=d?d@dAZ*ej)dBd>dCd?dDdAZ+ej)d>dEdFdGdHdAZ,ej)d d5d d ddIdJZ-W d n1 sw Y ej.dKdLdMdNZ/W d n 1 sw Y ej#d d0 ej0dOdPdQdRZ1ej$dSd8dTd5dUZ2W d n 1 s=w Y W d n 1 sMw Y ej3g dVg dWg dXg dYge%e*e+e,gdZd[ e/j4e e%e*e+e,e-e'ge1e2gd\ e d] W d n 1 sw Y ddd_d`Z5dadb Z6e7dckre6 dS dS )ez&
Gradio Web Interface for Soprano TTS
N)
SopranoTTS) play_streamz#Soprano Text-to-Speech Gradio WebUI) descriptionz --model-pathz-mz(Path to local model directory (optional))helpz--devicez-dauto)r cudacpumpszDevice to use for inference)defaultchoicesr z --backendz-b)r transformerslmdeployz Backend to use for inferencez --cache-sizez-cd z'Cache size in MB (for lmdeploy backend))typer
r z--decoder-batch-sizez-bs z Batch size when decoding audioz Loading Soprano TTS model...)backenddevice
cache_size_mbdecoder_batch_size
model_pathzModel loaded successfully!i } text temperaturetop_prepetition_penalty
chunk_size streamingc
c s, | s
dV d S zn|r-tj| ||||d}dV t|}d d|d ddfV W d S t }tj| |||d} t | }
| } | d t
j } t | t
}
|
d
kr\|
|
ntd }d |
dd
|
dd|dd}t
| f|fV W d S ty } zd dt| fV W Y d }~d S d }~ww )N)Nz*Please enter some text to generate speech.)r r r r )Nu ⏳ Streaming...u ✓ Streaming complete | i z.2fz ms latency)r r r i r infu ✓ Generated z s audio | Generation time: z.3fz s (z x realtime)u ✗ Error: )stripmodel infer_streamr time perf_counterinferr numpyastypenpint16len SAMPLE_RATEfloat Exceptionstr)r r r r r r streamlatency
start_timeaudiogen_timeaudio_np audio_int16
audio_secondsrtfstatuse r7 A/home/ubuntu/.local/lib/python3.10/site-packages/soprano/webui.pygenerate_speech. sZ
"r9 z Soprano TTS)titleu # 🗣️ Soprano TTS
**Device:** z | **Backend:** z
**Model Weights:** https://huggingface.co/ekwek/Soprano-1.1-80M
**Model Demo:** https://huggingface.co/spaces/ekwek/Soprano-TTS
**GitHub:** https://github.com/ekwek1/soprano
)scalezText to SynthesizezEnter text here...z|Soprano is an extremely lightweight text to speech model designed to produce highly realistic speech at unprecedented speed.
)label placeholdervaluelines max_linesz Stream AudioFzTNote: This bypasses the Gradio interface and streams audio directly to your speaker.)r? rA infozAdvanced Settings)open g ?g? Temperature)minimummaximumrA stepr? g ?ffffff?zTop Pg @333333?g?zRepetition PenaltyzChunk Size (Streaming only))rH rI rA rJ precisionr? zGenerate Speechprimarylg)variantsizezGenerated Speechr# T)r? r autoplayStatus )r? interactiverB rC )z9Soprano is an extremely lightweight text to speech model.rF rK rL )z2Artificial intelligence is transforming the world.rF rK rL )z"I'm so excited, I can't even wait!rF rK rL )z"Why don't you go ahead and try it?rF rK rL zExample Prompts)examplesinputsr? )fnrW outputsa
### Usage tips:
- When quoting, use double quotes instead of single quotes.
- Soprano works best when each sentence is between 2 and 30 seconds long.
- Although Soprano recognizes numbers and some special characters, it occasionally mispronounces them.
Best results can be achieved by converting these into their phonetic form.
(1+1 -> one plus one, etc)
- If Soprano produces unsatisfactory results, you can easily regenerate it for a new, potentially better generation.
You may also change the sampling settings for more varied results.
c C sz t | | | D ]1}z&ttjtj}|d|f |W d W S 1 s)w Y W q ty8 Y qw td)N zCould not find a free port)rangesocketAF_INET SOCK_STREAMbindOSError)
start_port max_triesportsr7 r7 r8 find_free_port s * rf c C s8 t d} td| tjd| dtjjdddd d S ) NrZ z"Starting Gradio interface on port z 127.0.0.1Fgreen) primary_huezQ
a {
color: var(--primary-600);
}
a:hover {
color: var(--primary-700);
}
) server_name server_portsharethemecss)rf printdemolaunchgrthemesSoft)rd r7 r7 r8 main s
rt __main__)rZ r )8__doc__argparser] r gradiorq r# r% sopranor soprano.utils.streamingr ArgumentParserparser add_argumentint
parse_argsargsrn r r
cache_sizer r r r( r+ r) boolr9 Blocksro MarkdownupperRowColumnTextbox
text_inputCheckboxr AccordionSliderr r r r Button generate_btnAudio audio_output
status_outputExamplesclickrf rt __name__r7 r7 r7 r8 s"
>
, 9
f