|
| 1 | +package main |
| 2 | + |
| 3 | +import ( |
| 4 | + "bytes" |
| 5 | + "context" |
| 6 | + "errors" |
| 7 | + "fmt" |
| 8 | + "io" |
| 9 | + "regexp" |
| 10 | + "strings" |
| 11 | + |
| 12 | + // Packages |
| 13 | + tablewriter "github.com/djthorpe/go-tablewriter" |
| 14 | + audio "github.com/go-audio/audio" |
| 15 | + wav "github.com/go-audio/wav" |
| 16 | + client "github.com/mutablelogic/go-client" |
| 17 | + elevenlabs "github.com/mutablelogic/go-client/pkg/elevenlabs" |
| 18 | + |
| 19 | + // Namespace imports |
| 20 | + . "github.com/djthorpe/go-errors" |
| 21 | +) |
| 22 | + |
| 23 | +/////////////////////////////////////////////////////////////////////////////// |
| 24 | +// GLOBALS |
| 25 | + |
| 26 | +var ( |
| 27 | + elName = "elevenlabs" |
| 28 | + elClient *elevenlabs.Client |
| 29 | + elExt = "mp3" |
| 30 | + elBitrate = uint64(32) // in kbps |
| 31 | + elSamplerate = uint64(44100) // in Hz |
| 32 | + elSimilarityBoost = float64(0.0) |
| 33 | + elStability = float64(0.0) |
| 34 | + elUseSpeakerBoost = false |
| 35 | + elWriteSettings = false |
| 36 | + reVoiceId = regexp.MustCompile("^[a-z0-9-]{20}$") |
| 37 | +) |
| 38 | + |
| 39 | +/////////////////////////////////////////////////////////////////////////////// |
| 40 | +// LIFECYCLE |
| 41 | + |
| 42 | +func elRegister(flags *Flags) { |
| 43 | + // Register flags required |
| 44 | + flags.String(elName, "elevenlabs-api-key", "${ELEVENLABS_API_KEY}", "API Key") |
| 45 | + flags.Float(elName, "similarity-boost", 0, "Similarity boost") |
| 46 | + flags.Float(elName, "stability", 0, "Voice stability") |
| 47 | + flags.Bool(elName, "use-speaker-boost", false, "Use speaker boost") |
| 48 | + flags.Unsigned(elName, "bitrate", 0, "Bit rate (kbps)") |
| 49 | + flags.Unsigned(elName, "samplerate", 0, "Sample rate (kHz)") |
| 50 | + |
| 51 | + // Register command set |
| 52 | + flags.Register(Cmd{ |
| 53 | + Name: elName, |
| 54 | + Description: "Elevenlabs API", |
| 55 | + Parse: elParse, |
| 56 | + Fn: []Fn{ |
| 57 | + {Name: "models", Call: elModels, Description: "Gets a list of available models"}, |
| 58 | + {Name: "voices", Call: elVoices, Description: "Return registered voices"}, |
| 59 | + {Name: "voice", Call: elVoice, Description: "Return one voice", MinArgs: 1, MaxArgs: 1, Syntax: "<voice-id>"}, |
| 60 | + {Name: "settings", Call: elVoiceSettings, Description: "Return voice settings, or default settings. Set voice settings from -stability, -similarity-boost and -use-speaker-boost flags", MaxArgs: 1, Syntax: "(<voice-id>)"}, |
| 61 | + {Name: "say", Call: elTextToSpeech, Description: "Text to speech", MinArgs: 2, Syntax: "<voice-id> <text>..."}, |
| 62 | + }, |
| 63 | + }) |
| 64 | +} |
| 65 | + |
| 66 | +func elParse(flags *Flags, opts ...client.ClientOpt) error { |
| 67 | + // Set defaults |
| 68 | + if typ := flags.GetOutExt(); typ != "" { |
| 69 | + elExt = strings.ToLower(flags.GetOutExt()) |
| 70 | + } |
| 71 | + |
| 72 | + // Create the client |
| 73 | + apiKey := flags.GetString("elevenlabs-api-key") |
| 74 | + if apiKey == "" { |
| 75 | + return fmt.Errorf("missing -elevenlabs-api-key flag") |
| 76 | + } else if client, err := elevenlabs.New(apiKey, opts...); err != nil { |
| 77 | + return err |
| 78 | + } else { |
| 79 | + elClient = client |
| 80 | + } |
| 81 | + |
| 82 | + // Get the bit rate and sample rate |
| 83 | + if bitrate, err := flags.GetValue("bitrate"); err == nil { |
| 84 | + if bitrate_, ok := bitrate.(uint64); ok && bitrate_ > 0 { |
| 85 | + elBitrate = bitrate_ |
| 86 | + } |
| 87 | + } |
| 88 | + if samplerate, err := flags.GetValue("samplerate"); err == nil { |
| 89 | + if samplerate_, ok := samplerate.(uint64); ok && samplerate_ > 0 { |
| 90 | + elSamplerate = samplerate_ |
| 91 | + } |
| 92 | + } |
| 93 | + |
| 94 | + // Similarity boost |
| 95 | + if value, err := flags.GetValue("similarity-boost"); err == nil { |
| 96 | + elSimilarityBoost = value.(float64) |
| 97 | + elWriteSettings = true |
| 98 | + } else if !errors.Is(err, ErrNotFound) { |
| 99 | + return err |
| 100 | + } |
| 101 | + |
| 102 | + // Stability |
| 103 | + if value, err := flags.GetValue("stability"); err == nil { |
| 104 | + elStability = value.(float64) |
| 105 | + elWriteSettings = true |
| 106 | + } else if !errors.Is(err, ErrNotFound) { |
| 107 | + return err |
| 108 | + } |
| 109 | + |
| 110 | + // Use speaker boost |
| 111 | + if value, err := flags.GetValue("use-speaker-boost"); err == nil { |
| 112 | + elUseSpeakerBoost = value.(bool) |
| 113 | + elWriteSettings = true |
| 114 | + } else if !errors.Is(err, ErrNotFound) { |
| 115 | + return err |
| 116 | + } |
| 117 | + |
| 118 | + // Return success |
| 119 | + return nil |
| 120 | +} |
| 121 | + |
| 122 | +///////////////////////////////////////////////////////////////////// |
| 123 | +// API CALL FUNCTIONS |
| 124 | + |
| 125 | +func elModels(ctx context.Context, w *tablewriter.Writer, args []string) error { |
| 126 | + models, err := elClient.Models() |
| 127 | + if err != nil { |
| 128 | + return err |
| 129 | + } |
| 130 | + return w.Write(models) |
| 131 | +} |
| 132 | + |
| 133 | +func elVoices(ctx context.Context, w *tablewriter.Writer, args []string) error { |
| 134 | + voices, err := elClient.Voices() |
| 135 | + if err != nil { |
| 136 | + return err |
| 137 | + } |
| 138 | + return w.Write(voices) |
| 139 | +} |
| 140 | + |
| 141 | +func elVoice(ctx context.Context, w *tablewriter.Writer, args []string) error { |
| 142 | + if voice, err := elVoiceId(args[0]); err != nil { |
| 143 | + return err |
| 144 | + } else if voice, err := elClient.Voice(voice); err != nil { |
| 145 | + return err |
| 146 | + } else { |
| 147 | + return w.Write(voice) |
| 148 | + } |
| 149 | +} |
| 150 | + |
| 151 | +func elVoiceSettings(ctx context.Context, w *tablewriter.Writer, args []string) error { |
| 152 | + var voice string |
| 153 | + if len(args) > 0 { |
| 154 | + if v, err := elVoiceId(args[0]); err != nil { |
| 155 | + return err |
| 156 | + } else { |
| 157 | + voice = v |
| 158 | + } |
| 159 | + } |
| 160 | + |
| 161 | + // Get voice settings |
| 162 | + settings, err := elClient.VoiceSettings(voice) |
| 163 | + if err != nil { |
| 164 | + return err |
| 165 | + } |
| 166 | + |
| 167 | + // Modify settings |
| 168 | + if elWriteSettings { |
| 169 | + // We need a voice in order to write the settings |
| 170 | + if voice == "" { |
| 171 | + return ErrBadParameter.With("Missing voice-id") |
| 172 | + } |
| 173 | + |
| 174 | + // Change parameters |
| 175 | + if elStability != 0.0 { |
| 176 | + settings.Stability = float32(elStability) |
| 177 | + } |
| 178 | + if elSimilarityBoost != 0.0 { |
| 179 | + settings.SimilarityBoost = float32(elSimilarityBoost) |
| 180 | + } |
| 181 | + if elUseSpeakerBoost != settings.UseSpeakerBoost { |
| 182 | + settings.UseSpeakerBoost = elUseSpeakerBoost |
| 183 | + } |
| 184 | + |
| 185 | + // Set voice settings |
| 186 | + if err := elClient.SetVoiceSettings(voice, settings); err != nil { |
| 187 | + return err |
| 188 | + } |
| 189 | + } |
| 190 | + |
| 191 | + return w.Write(settings) |
| 192 | +} |
| 193 | + |
| 194 | +func elTextToSpeech(ctx context.Context, w *tablewriter.Writer, args []string) error { |
| 195 | + // The voice to use |
| 196 | + voice, err := elVoiceId(args[0]) |
| 197 | + if err != nil { |
| 198 | + return err |
| 199 | + } |
| 200 | + |
| 201 | + // Output format |
| 202 | + opts := []elevenlabs.Opt{} |
| 203 | + if format := elOutputFormat(); format != nil { |
| 204 | + opts = append(opts, format) |
| 205 | + } else { |
| 206 | + return ErrBadParameter.Withf("invalid output format %q", elExt) |
| 207 | + } |
| 208 | + |
| 209 | + // The text to speak |
| 210 | + text := strings.Join(args[1:], " ") |
| 211 | + |
| 212 | + // If wav, then wrap in a header |
| 213 | + if elExt == "wav" { |
| 214 | + // Create the writer |
| 215 | + writer := NewAudioWriter(w.Output().(io.WriteSeeker), int(elSamplerate), 1) |
| 216 | + defer writer.Close() |
| 217 | + |
| 218 | + // Read the data |
| 219 | + if n, err := elClient.TextToSpeech(writer, voice, text, opts...); err != nil { |
| 220 | + return err |
| 221 | + } else { |
| 222 | + elClient.Debugf("elTextToSpeech: generated %v bytes of PCM data", n) |
| 223 | + } |
| 224 | + } else if _, err := elClient.TextToSpeech(w.Output(), voice, text, opts...); err != nil { |
| 225 | + return err |
| 226 | + } |
| 227 | + |
| 228 | + // Return success |
| 229 | + return nil |
| 230 | +} |
| 231 | + |
| 232 | +///////////////////////////////////////////////////////////////////// |
| 233 | +// PRIVATE METHODS |
| 234 | + |
| 235 | +func elVoiceId(q string) (string, error) { |
| 236 | + if reVoiceId.MatchString(q) { |
| 237 | + return q, nil |
| 238 | + } else if voices, err := elClient.Voices(); err != nil { |
| 239 | + return "", err |
| 240 | + } else { |
| 241 | + for _, v := range voices { |
| 242 | + if strings.EqualFold(v.Name, q) || v.Id == q { |
| 243 | + return v.Id, nil |
| 244 | + } |
| 245 | + } |
| 246 | + } |
| 247 | + return "", ErrNotFound.Withf("%q", q) |
| 248 | +} |
| 249 | + |
| 250 | +func elOutputFormat() elevenlabs.Opt { |
| 251 | + switch elExt { |
| 252 | + case "mp3": |
| 253 | + return elevenlabs.OptFormatMP3(uint(elBitrate), uint(elSamplerate)) |
| 254 | + case "wav": |
| 255 | + return elevenlabs.OptFormatPCM(uint(elSamplerate)) |
| 256 | + case "ulaw": |
| 257 | + return elevenlabs.OptFormatULAW() |
| 258 | + } |
| 259 | + return nil |
| 260 | +} |
| 261 | + |
| 262 | +///////////////////////////////////////////////////////////////////// |
| 263 | +// AUDIO WRITER |
| 264 | + |
| 265 | +type wavWriter struct { |
| 266 | + enc *wav.Encoder |
| 267 | + buf *bytes.Buffer |
| 268 | + pcm *audio.IntBuffer |
| 269 | +} |
| 270 | + |
| 271 | +func NewAudioWriter(w io.WriteSeeker, sampleRate, channels int) *wavWriter { |
| 272 | + this := new(wavWriter) |
| 273 | + |
| 274 | + // Create a WAV encoder |
| 275 | + this.enc = wav.NewEncoder(w, sampleRate, 16, channels, 1) |
| 276 | + if this.enc == nil { |
| 277 | + return nil |
| 278 | + } |
| 279 | + |
| 280 | + // Create a buffer for the incoming byte data |
| 281 | + this.buf = bytes.NewBuffer(nil) |
| 282 | + |
| 283 | + // Make a PCM buffer with a capacity of 4096 samples |
| 284 | + this.pcm = &audio.IntBuffer{ |
| 285 | + Format: &audio.Format{ |
| 286 | + SampleRate: this.enc.SampleRate, |
| 287 | + NumChannels: this.enc.NumChans, |
| 288 | + }, |
| 289 | + SourceBitDepth: this.enc.BitDepth, |
| 290 | + Data: make([]int, 0, 4096), |
| 291 | + } |
| 292 | + |
| 293 | + // Return the writer |
| 294 | + return this |
| 295 | +} |
| 296 | + |
| 297 | +func (a *wavWriter) Write(data []byte) (int, error) { |
| 298 | + // Write the data to the buffer |
| 299 | + if n, err := a.buf.Write(data); err != nil { |
| 300 | + return 0, err |
| 301 | + } else if err := a.Flush(); err != nil { |
| 302 | + return 0, err |
| 303 | + } else { |
| 304 | + return n, nil |
| 305 | + } |
| 306 | +} |
| 307 | + |
| 308 | +func (a *wavWriter) Flush() error { |
| 309 | + var n int |
| 310 | + var sample [2]byte |
| 311 | + |
| 312 | + // Read data until we have a full PCM buffer |
| 313 | + for { |
| 314 | + if a.buf.Len() < len(sample) { |
| 315 | + break |
| 316 | + } else if n, err := a.buf.Read(sample[:]); err != nil { |
| 317 | + return err |
| 318 | + } else if n != len(sample) { |
| 319 | + return ErrInternalAppError.With("short read") |
| 320 | + } |
| 321 | + |
| 322 | + // Append the sample data - Little Endian |
| 323 | + a.pcm.Data = append(a.pcm.Data, int(int16(sample[0])|int16(sample[1])<<8)) |
| 324 | + n += 2 |
| 325 | + } |
| 326 | + |
| 327 | + // Write the PCM data |
| 328 | + if n > 0 { |
| 329 | + if err := a.enc.Write(a.pcm); err != nil { |
| 330 | + return err |
| 331 | + } |
| 332 | + } |
| 333 | + |
| 334 | + // Reset the PCM data |
| 335 | + a.pcm.Data = a.pcm.Data[:0] |
| 336 | + |
| 337 | + // Return success |
| 338 | + return nil |
| 339 | +} |
| 340 | + |
| 341 | +func (a *wavWriter) Close() error { |
| 342 | + if err := a.Flush(); err != nil { |
| 343 | + return err |
| 344 | + } |
| 345 | + return a.enc.Close() |
| 346 | +} |
0 commit comments