add elevenlabs tts

README.md

@@ -88,7 +88,7 @@ providers:
 ```yaml
 providers:
   - type: cohere
-    token: ${COHERE_API_TOKEN}
+    token: ${COHERE_API_KEY}
 
     # https://docs.cohere.com/docs/models
     models:
@@ -105,7 +105,7 @@ providers:
 ```yaml
 providers:
   - type: groq
-    token: ${GROQ_API_TOKEN}
+    token: ${GROQ_API_KEY}
 
     # https://console.groq.com/docs/models
     models:
@@ -122,7 +122,7 @@ providers:
 ```yaml
 providers:
   - type: mistral
-    token: ${MISTRAL_API_TOKEN}
+    token: ${MISTRAL_API_KEY}
 
     # https://docs.mistral.ai/getting-started/models/
     models:
@@ -138,7 +138,7 @@ https://replicate.com/
 ```yaml
 providers:
   - type: replicate
-    token: ${REPLICATE_API_TOKEN}
+    token: ${REPLICATE_API_KEY}
     models:
       replicate-flux-pro:
         id: black-forest-labs/flux-pro
@@ -262,6 +262,22 @@ providers:
 ```
 
 
+#### Eleven Labs
+
+```yaml
+providers:
+  - type: elevenlabs
+    token: ${ELEVENLABS_API_KEY}
+
+    models:
+      elevenlabs-sarah:
+        id: EXAVITQu4vr4xnSDxMaL
+
+      elevenlabs-charlie:
+        id: IKne3meq5aSn9XLyUdCD
+```
+
+
 #### Mimic 3
 
 ```shell

cmd/client/main.go

@@ -273,7 +273,11 @@ LOOP:
 			continue LOOP
 		}
 
-		name := uuid.New().String() + ".wav"
+		name := uuid.New().String()
+
+		if ext, _ := mime.ExtensionsByType(http.DetectContentType(data)); len(ext) > 0 {
+			name += ext[0]
+		}
 
 		os.WriteFile(name, data, 0600)
 		fmt.Println("Saved: " + name)

config/config_model.go

@@ -123,6 +123,7 @@ func DetectModelType(id string) ModelType {
 	}
 
 	synthesizers := []string{
+		"eleven",
 		"stable-audio",
 		"tts",
 	}

config/config_provider.go

@@ -9,6 +9,10 @@ import (
 func (cfg *Config) registerProviders(f *configFile) error {
 	for _, p := range f.Providers {
 		for id, m := range p.Models {
+			if m.Type == "" {
+				m.Type = DetectModelType(m.ID)
+			}
+
 			if m.Type == "" {
 				m.Type = DetectModelType(id)
 			}

config/config_synthesizer.go

@@ -6,6 +6,7 @@ import (
 
 	"github.com/adrianliechti/llama/pkg/provider"
 	"github.com/adrianliechti/llama/pkg/provider/coqui"
+	"github.com/adrianliechti/llama/pkg/provider/elevenlabs"
 	"github.com/adrianliechti/llama/pkg/provider/mimic"
 	"github.com/adrianliechti/llama/pkg/provider/openai"
 
@@ -43,6 +44,9 @@ func createSynthesizer(cfg providerConfig, model modelContext) (provider.Synthes
 	case "coqui":
 		return coquiSynthesizer(cfg, model)
 
+	case "elevenlabs":
+		return elevenlabsSynthesizer(cfg, model)
+
 	case "mimic":
 		return mimicSynthesizer(cfg, model)
 
@@ -60,6 +64,24 @@ func coquiSynthesizer(cfg providerConfig, model modelContext) (provider.Synthesi
 	return coqui.NewSynthesizer(cfg.URL, options...)
 }
 
+func elevenlabsSynthesizer(cfg providerConfig, model modelContext) (provider.Synthesizer, error) {
+	var options []elevenlabs.Option
+
+	if cfg.URL != "" {
+		options = append(options, elevenlabs.WithURL(cfg.URL))
+	}
+
+	if cfg.Token != "" {
+		options = append(options, elevenlabs.WithToken(cfg.Token))
+	}
+
+	if model.ID != "" {
+		options = append(options, elevenlabs.WithModel(model.ID))
+	}
+
+	return elevenlabs.NewSynthesizer(options...)
+}
+
 func mimicSynthesizer(cfg providerConfig, model modelContext) (provider.Synthesizer, error) {
 	var options []mimic.Option
 

pkg/provider/elevenlabs/config.go

@@ -0,0 +1,40 @@
+package elevenlabs
+
+import (
+	"net/http"
+)
+
+type Config struct {
+	url string
+
+	token string
+	model string
+
+	client *http.Client
+}
+
+type Option func(*Config)
+
+func WithClient(client *http.Client) Option {
+	return func(c *Config) {
+		c.client = client
+	}
+}
+
+func WithURL(url string) Option {
+	return func(c *Config) {
+		c.url = url
+	}
+}
+
+func WithToken(token string) Option {
+	return func(c *Config) {
+		c.token = token
+	}
+}
+
+func WithModel(model string) Option {
+	return func(c *Config) {
+		c.model = model
+	}
+}

pkg/provider/elevenlabs/synthesizer.go

@@ -0,0 +1,69 @@
+package elevenlabs
+
+import (
+	"context"
+	"net/http"
+	"net/url"
+	"strings"
+
+	"github.com/adrianliechti/llama/pkg/provider"
+
+	"github.com/google/uuid"
+)
+
+var _ provider.Synthesizer = (*Synthesizer)(nil)
+
+type Synthesizer struct {
+	*Config
+}
+
+func NewSynthesizer(options ...Option) (*Synthesizer, error) {
+	cfg := &Config{
+		url: "https://api.elevenlabs.io",
+
+		client: http.DefaultClient,
+	}
+
+	for _, option := range options {
+		option(cfg)
+	}
+
+	return &Synthesizer{
+		Config: cfg,
+	}, nil
+}
+
+func (s *Synthesizer) Synthesize(ctx context.Context, content string, options *provider.SynthesizeOptions) (*provider.Synthesis, error) {
+	if options == nil {
+		options = new(provider.SynthesizeOptions)
+	}
+
+	u, _ := url.Parse(strings.TrimRight(s.url, "/") + "/v1/text-to-speech/" + s.model)
+
+	body := map[string]any{
+		"text":     content,
+		"model_id": "eleven_multilingual_v2",
+	}
+
+	req, _ := http.NewRequestWithContext(ctx, "POST", u.String(), jsonReader(body))
+	req.Header.Set("xi-api-key", s.token)
+
+	resp, err := s.client.Do(req)
+
+	if err != nil {
+		return nil, err
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		return nil, convertError(resp)
+	}
+
+	id := uuid.NewString()
+
+	return &provider.Synthesis{
+		ID: id,
+
+		Name:    id + ".mp3",
+		Content: resp.Body,
+	}, nil
+}

pkg/provider/elevenlabs/util.go

@@ -0,0 +1,29 @@
+package elevenlabs
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"io"
+	"net/http"
+)
+
+func convertError(resp *http.Response) error {
+	data, _ := io.ReadAll(resp.Body)
+
+	if len(data) == 0 {
+		return errors.New(http.StatusText(resp.StatusCode))
+	}
+
+	return errors.New(string(data))
+}
+
+func jsonReader(v any) io.Reader {
+	b := new(bytes.Buffer)
+
+	enc := json.NewEncoder(b)
+	enc.SetEscapeHTML(false)
+
+	enc.Encode(v)
+	return b
+}