File size: 5,453 Bytes
25647ae
 
79eafc9
25647ae
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
79eafc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25647ae
 
79eafc9
25647ae
 
 
 
 
 
 
 
 
79eafc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25647ae
 
 
 
79eafc9
25647ae
 
 
 
 
 
 
 
 
 
 
 
 
79eafc9
 
 
 
 
25647ae
79eafc9
 
 
 
 
 
25647ae
79eafc9
 
25647ae
79eafc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25647ae
 
 
79eafc9
25647ae
 
 
 
 
 
 
 
 
 
79eafc9
 
25647ae
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
/* eslint-disable no-restricted-globals */
import { pipeline } from 'https://cdn.jsdelivr.net/npm/@huggingface/transformers@latest'
import { KokoroTTS } from 'https://cdn.jsdelivr.net/npm/kokoro-js@1.2.1/dist/kokoro.web.js'

class MyTextToSpeechPipeline {
  static task = 'text-to-speech'
  static instance = null

  static async getInstance(model, dtype = 'fp32', progress_callback = null) {
    try {
      // Try WebGPU first
      this.instance = await pipeline(this.task, model, {
        dtype,
        device: 'webgpu',
        progress_callback,
        quantized: false
      })
      return this.instance
    } catch (webgpuError) {
      // Fallback to WASM if WebGPU fails
      if (progress_callback) {
        progress_callback({
          status: 'fallback',
          message: 'WebGPU failed, falling back to WASM'
        })
      }
      try {
        this.instance = await pipeline(this.task, model, {
          dtype,
          device: 'wasm',
          progress_callback,
          quantized: false
        })
        return this.instance
      } catch (wasmError) {
        throw new Error(
          `Both WebGPU and WASM failed. WebGPU error: ${webgpuError.message}. WASM error: ${wasmError.message}`
        )
      }
    }
  }
}

class MyKokoroTTSPipeline {
  static instance = null

  static async getInstance(model, dtype = 'fp32', progress_callback = null) {
    try {
      const device = 'webgpu'
      if (progress_callback) {
        progress_callback({
          status: 'loading',
          message: `Loading Kokoro TTS model with ${device} device`
        })
      }

      this.instance = await KokoroTTS.from_pretrained(model, {
        dtype,
        device,
        progress_callback: progress_callback
          ? (data) => {
              progress_callback({
                status: 'loading',
                ...data
              })
            }
          : null
      })
      return this.instance
    } catch (webgpuError) {
      // Fallback to WASM if WebGPU fails
      if (progress_callback) {
        progress_callback({
          status: 'fallback',
          message: 'WebGPU failed, falling back to WASM'
        })
      }
      try {
        this.instance = await KokoroTTS.from_pretrained(model, {
          dtype,
          device: 'wasm',
          progress_callback: progress_callback
            ? (data) => {
                progress_callback({
                  status: 'loading',
                  ...data
                })
              }
            : null
        })
        return this.instance
      } catch (wasmError) {
        throw new Error(
          `Both WebGPU and WASM failed for Kokoro TTS. WebGPU error: ${webgpuError.message}. WASM error: ${wasmError.message}`
        )
      }
    }
  }
}

self.addEventListener('message', async (event) => {
  try {
    const { type, model, dtype, text, isStyleTTS2, config } = event.data

    if (!model) {
      self.postMessage({
        status: 'error',
        output: 'No model provided'
      })
      return
    }

    let synthesizer
    if (isStyleTTS2) {
      // Use Kokoro TTS for StyleTTS2 models
      synthesizer = await MyKokoroTTSPipeline.getInstance(
        model,
        dtype || 'q8',
        (x) => {
          self.postMessage({ status: 'loading', output: x })
        }
      )
    } else {
      // Use standard transformers pipeline
      synthesizer = await MyTextToSpeechPipeline.getInstance(
        model,
        dtype || 'fp32',
        (x) => {
          self.postMessage({ status: 'loading', output: x })
        }
      )
    }

    if (type === 'load') {
      self.postMessage({
        status: 'ready',
        output: `Model ${model}${isStyleTTS2 ? ' StyleTTS2' : ''}, dtype ${dtype} loaded`
      })
      return
    }

    if (type === 'synthesize') {
      if (!text || typeof text !== 'string' || text.trim() === '') {
        self.postMessage({
          status: 'error',
          output: 'No text provided for synthesis'
        })
        return
      }

      try {
        let output

        if (isStyleTTS2) {
          const options = {}

          options.voice = config.voice
          const audioResult = await synthesizer.generate(text.trim(), options)

          output = {
            audio: Array.from(audioResult.audio),
            sampling_rate: audioResult.sampling_rate || 24000 // Default for Kokoro
          }
        } else {
          const options = {}

          if (config?.speakerEmbeddings) {
            try {
              const response = await fetch(config.speakerEmbeddings)
              if (response.ok) {
                const embeddings = await response.arrayBuffer()
                options.speaker_embeddings = new Float32Array(embeddings)
              }
            } catch (error) {
              console.warn('Failed to load speaker embeddings:', error)
            }
          }

          const result = await synthesizer(text.trim(), options)
          output = {
            audio: Array.from(result.audio),
            sampling_rate: result.sampling_rate
          }
        }

        self.postMessage({
          status: 'output',
          output
        })

        self.postMessage({ status: 'ready' })
      } catch (error) {
        throw error
      }
    }
  } catch (error) {
    self.postMessage({
      status: 'error',
      output:
        error.message || 'An error occurred during text-to-speech synthesis'
    })
  }
})