talk/internal/provider/catalog.go

package provider

import "sort"

// Voice is an entry of the voice catalog: a known voice of a provider for a
// language, with a readable name. The catalog is curated (the common
// standard voices), not complete; any voice ID of the provider also works
// without a catalog entry (--voice, config).
type Voice struct {
	Provider string
	Lang     string // de | en
	ID       string
	Name     string
}

// catalog lists the known voices per provider. ElevenLabs: public premade
// voices (multilingual, hence listed for both de AND en). OpenAI: the voices
// of the TTS API (language independent). Edge: neural voices of the
// respective locale. osbuiltin has no IDs (the OS chooses).
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 returns the voice catalog in stable order (chain order of the
// providers, then de before en, then catalog order).
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 returns the catalog name of a voice ID; empty if unknown.
func VoiceName(provider, id string) string {
	for _, v := range catalog {
		if v.Provider == provider && v.ID == id {
			return v.Name
		}
	}
	return ""
}

// DefaultVoice returns the effective default voice of a provider for a
// language: config mapping voice.<provider>.<lang> before the repo default
// (without a --voice pin). Empty for osbuiltin (the OS chooses).
func DefaultVoice(provider, lang string, cfgVoices map[string]map[string]string) string {
	return resolveVoice(provider, lang, "", cfgVoices[provider])
}