Files
talk/internal/provider/catalog.go
T

90 lines
3.4 KiB
Go
Raw Normal View History

package provider
import "sort"
// Voice ist ein Eintrag des Stimmen-Katalogs: eine bekannte Stimme eines
// Providers für eine Sprache, mit lesbarem Namen. Der Katalog ist kuratiert
// (die gängigen Standardstimmen), nicht vollständig — jede Voice-ID des
// Providers funktioniert auch ohne Katalog-Eintrag (--voice, config).
type Voice struct {
Provider string
Lang string // de | en
ID string
Name string
}
// catalog sind die bekannten Stimmen je Provider. ElevenLabs: öffentliche
// Premade-Stimmen (multilingual, daher für de UND en gelistet). OpenAI: die
// Stimmen der TTS-API (sprachunabhängig). Edge: Neural-Stimmen der
// jeweiligen Locale. osbuiltin kennt keine IDs (das OS wählt).
var catalog = []Voice{
{"elevenlabs", "de", "21m00Tcm4TlvDq8ikWAM", "Rachel"},
{"elevenlabs", "de", "pNInz6obpgDQGcFmaJgB", "Adam"},
{"elevenlabs", "de", "EXAVITQu4vr4xnSDxMaL", "Bella"},
{"elevenlabs", "de", "ErXwobaYiN019PkySvjV", "Antoni"},
{"elevenlabs", "de", "TxGEqnHWrfWFTfGW9XjX", "Josh"},
{"elevenlabs", "en", "21m00Tcm4TlvDq8ikWAM", "Rachel"},
{"elevenlabs", "en", "pNInz6obpgDQGcFmaJgB", "Adam"},
{"elevenlabs", "en", "EXAVITQu4vr4xnSDxMaL", "Bella"},
{"elevenlabs", "en", "ErXwobaYiN019PkySvjV", "Antoni"},
{"elevenlabs", "en", "TxGEqnHWrfWFTfGW9XjX", "Josh"},
{"openai", "de", "alloy", "Alloy"},
{"openai", "de", "echo", "Echo"},
{"openai", "de", "fable", "Fable"},
{"openai", "de", "onyx", "Onyx"},
{"openai", "de", "nova", "Nova"},
{"openai", "de", "shimmer", "Shimmer"},
{"openai", "en", "alloy", "Alloy"},
{"openai", "en", "echo", "Echo"},
{"openai", "en", "fable", "Fable"},
{"openai", "en", "onyx", "Onyx"},
{"openai", "en", "nova", "Nova"},
{"openai", "en", "shimmer", "Shimmer"},
{"edge", "de", "de-DE-KatjaNeural", "Katja"},
{"edge", "de", "de-DE-ConradNeural", "Conrad"},
{"edge", "de", "de-DE-AmalaNeural", "Amala"},
{"edge", "de", "de-DE-KillianNeural", "Killian"},
{"edge", "de", "de-AT-IngridNeural", "Ingrid (AT)"},
{"edge", "de", "de-CH-LeniNeural", "Leni (CH)"},
{"edge", "en", "en-US-AriaNeural", "Aria"},
{"edge", "en", "en-US-GuyNeural", "Guy"},
{"edge", "en", "en-US-JennyNeural", "Jenny"},
{"edge", "en", "en-US-ChristopherNeural", "Christopher"},
{"edge", "en", "en-GB-SoniaNeural", "Sonia (GB)"},
{"edge", "en", "en-GB-RyanNeural", "Ryan (GB)"},
}
// Catalog liefert den Stimmen-Katalog in stabiler Reihenfolge
// (Ketten-Reihenfolge der Provider, dann de vor en, dann Katalog-Reihe).
func Catalog() []Voice {
out := make([]Voice, len(catalog))
copy(out, catalog)
rank := map[string]int{}
for i, name := range DefaultOrder {
rank[name] = i
}
sort.SliceStable(out, func(i, j int) bool {
if rank[out[i].Provider] != rank[out[j].Provider] {
return rank[out[i].Provider] < rank[out[j].Provider]
}
return out[i].Lang < out[j].Lang
})
return out
}
// VoiceName liefert den Katalog-Namen einer Voice-ID; leer, wenn unbekannt.
func VoiceName(provider, id string) string {
for _, v := range catalog {
if v.Provider == provider && v.ID == id {
return v.Name
}
}
return ""
}
// DefaultVoice liefert die effektive Default-Stimme eines Providers für
// eine Sprache: config-Mapping voice.<provider>.<lang> vor dem Repo-Default
// (ohne --voice-Pin). Leer bei osbuiltin (das OS wählt).
func DefaultVoice(provider, lang string, cfgVoices map[string]map[string]string) string {
return resolveVoice(provider, lang, "", cfgVoices[provider])
}