active_call/synthesis/
mod.rs

1use anyhow::Result;
2use async_trait::async_trait;
3use bytes::Bytes;
4use futures::stream::BoxStream;
5use serde::{Deserialize, Serialize};
6use std::collections::HashMap;
7use tokio::sync::mpsc;
8
9mod aliyun;
10mod deepgram;
11mod tencent_cloud;
12mod tencent_cloud_basic;
13
14#[cfg(feature = "offline")]
15mod supertonic;
16
17pub use aliyun::AliyunTtsClient;
18pub use deepgram::DeepegramTtsClient;
19pub use tencent_cloud::TencentCloudTtsClient;
20pub use tencent_cloud_basic::TencentCloudTtsBasicClient;
21
22#[cfg(feature = "offline")]
23pub use supertonic::SupertonicTtsClient;
24
25#[derive(Clone, Default)]
26pub struct SynthesisCommand {
27    pub text: String,
28    pub speaker: Option<String>,
29    pub play_id: Option<String>,
30    pub streaming: bool,
31    pub end_of_stream: bool,
32    pub option: SynthesisOption,
33    pub base64: bool,
34    pub cache_key: Option<String>,
35}
36pub type SynthesisCommandSender = mpsc::UnboundedSender<SynthesisCommand>;
37pub type SynthesisCommandReceiver = mpsc::UnboundedReceiver<SynthesisCommand>;
38
39#[derive(Debug, Clone, Serialize, Hash, Eq, PartialEq)]
40pub enum SynthesisType {
41    #[serde(rename = "tencent")]
42    TencentCloud,
43    #[serde(rename = "aliyun")]
44    Aliyun,
45    #[serde(rename = "deepgram")]
46    Deepgram,
47    #[cfg(feature = "offline")]
48    #[serde(rename = "supertonic")]
49    Supertonic,
50    #[serde(rename = "other")]
51    Other(String),
52}
53
54impl std::fmt::Display for SynthesisType {
55    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
56        match self {
57            SynthesisType::TencentCloud => write!(f, "tencent"),
58            SynthesisType::Aliyun => write!(f, "aliyun"),
59            SynthesisType::Deepgram => write!(f, "deepgram"),
60            #[cfg(feature = "offline")]
61            SynthesisType::Supertonic => write!(f, "supertonic"),
62            SynthesisType::Other(provider) => write!(f, "{}", provider),
63        }
64    }
65}
66
67impl<'de> Deserialize<'de> for SynthesisType {
68    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
69    where
70        D: serde::Deserializer<'de>,
71    {
72        let value = String::deserialize(deserializer)?;
73        match value.as_str() {
74            "tencent" => Ok(SynthesisType::TencentCloud),
75            "aliyun" => Ok(SynthesisType::Aliyun),
76            "deepgram" => Ok(SynthesisType::Deepgram),
77            #[cfg(feature = "offline")]
78            "supertonic" => Ok(SynthesisType::Supertonic),
79            _ => Ok(SynthesisType::Other(value)),
80        }
81    }
82}
83
84#[cfg(test)]
85mod tests;
86#[derive(Debug, Clone, Deserialize, Serialize)]
87#[serde(rename_all = "camelCase")]
88#[serde(default)]
89pub struct SynthesisOption {
90    pub samplerate: Option<i32>,
91    pub provider: Option<SynthesisType>,
92    pub speed: Option<f32>,
93    pub app_id: Option<String>,
94    pub secret_id: Option<String>,
95    #[serde(alias = "apiKey")]
96    pub secret_key: Option<String>,
97    pub volume: Option<i32>,
98    pub speaker: Option<String>,
99    pub codec: Option<String>,
100    pub subtitle: Option<bool>,
101    #[serde(alias = "voice")]
102    pub model: Option<String>,
103    pub language: Option<String>,
104    /// emotion: neutral、sad、happy、angry、fear、news、story、radio、poetry、
105    /// call、sajiao、disgusted、amaze、peaceful、exciting、aojiao、jieshuo
106    pub emotion: Option<String>,
107    pub endpoint: Option<String>,
108    pub extra: Option<HashMap<String, String>>,
109    pub max_concurrent_tasks: Option<usize>,
110    pub session_id: Option<String>,
111}
112
113impl SynthesisOption {
114    pub fn merge_with(&self, option: Option<SynthesisOption>) -> Self {
115        if let Some(other) = option {
116            Self {
117                samplerate: other.samplerate.or(self.samplerate),
118                provider: other.provider.or(self.provider.clone()),
119                speed: other.speed.or(self.speed),
120                app_id: other.app_id.or(self.app_id.clone()),
121                secret_id: other.secret_id.or(self.secret_id.clone()),
122                secret_key: other.secret_key.or(self.secret_key.clone()),
123                volume: other.volume.or(self.volume),
124                speaker: other.speaker.or(self.speaker.clone()),
125                codec: other.codec.or(self.codec.clone()),
126                subtitle: other.subtitle.or(self.subtitle),
127                model: other.model.or(self.model.clone()),
128                language: other.language.or(self.language.clone()),
129                emotion: other.emotion.or(self.emotion.clone()),
130                endpoint: other.endpoint.or(self.endpoint.clone()),
131                extra: other.extra.or(self.extra.clone()),
132                max_concurrent_tasks: other.max_concurrent_tasks.or(self.max_concurrent_tasks),
133                session_id: other.session_id.or(self.session_id.clone()),
134            }
135        } else {
136            self.clone()
137        }
138    }
139}
140
141#[derive(Debug)]
142pub enum SynthesisEvent {
143    /// Raw audio data chunk
144    AudioChunk(Bytes),
145    /// Progress information including completion status
146    Subtitles(Vec<Subtitle>),
147    Finished,
148}
149
150#[derive(Debug, Clone)]
151pub struct Subtitle {
152    pub text: String,
153    pub begin_time: u32,
154    pub end_time: u32,
155    pub begin_index: u32,
156    pub end_index: u32,
157}
158
159impl Subtitle {
160    pub fn new(
161        text: String,
162        begin_time: u32,
163        end_time: u32,
164        begin_index: u32,
165        end_index: u32,
166    ) -> Self {
167        Self {
168            text,
169            begin_time,
170            end_time,
171            begin_index,
172            end_index,
173        }
174    }
175}
176
177// calculate audio duration from bytes size and sample rate
178pub fn bytes_size_to_duration(bytes: usize, sample_rate: u32) -> u32 {
179    (500.0 * bytes as f32 / sample_rate as f32) as u32
180}
181
182#[async_trait]
183pub trait SynthesisClient: Send {
184    // provider of the synthesis client.
185    fn provider(&self) -> SynthesisType;
186
187    // connect to the synthesis service.
188    // (cmd_seq, result), return the cmd_seq that passed from `synthesize`
189    async fn start(
190        &mut self,
191    ) -> Result<BoxStream<'static, (Option<usize>, Result<SynthesisEvent>)>>;
192
193    // send text to the synthesis service.
194    // `cmd_seq` and `option` are used for non streaming mode
195    // for streaming mode, `cmd_seq` and `option` are None
196    async fn synthesize(
197        &mut self,
198        text: &str,
199        cmd_seq: Option<usize>,
200        option: Option<SynthesisOption>,
201    ) -> Result<()>;
202
203    async fn stop(&mut self) -> Result<()>;
204}
205
206impl Default for SynthesisOption {
207    fn default() -> Self {
208        Self {
209            samplerate: Some(16000),
210            provider: None,
211            speed: Some(1.0),
212            app_id: None,
213            secret_id: None,
214            secret_key: None,
215            volume: Some(5), // 0-10
216            speaker: None,
217            codec: Some("pcm".to_string()),
218            subtitle: None,
219            model: None,
220            language: None,
221            emotion: None,
222            endpoint: None,
223            extra: None,
224            max_concurrent_tasks: None,
225            session_id: None,
226        }
227    }
228}
229
230impl SynthesisOption {
231    pub fn check_default(&mut self) {
232        if let Some(provider) = &self.provider {
233            match provider.to_string().as_str() {
234                "tencent" | "tencent_basic" => {
235                    if self.app_id.is_none() {
236                        self.app_id = std::env::var("TENCENT_APPID").ok();
237                    }
238                    if self.secret_id.is_none() {
239                        self.secret_id = std::env::var("TENCENT_SECRET_ID").ok();
240                    }
241                    if self.secret_key.is_none() {
242                        self.secret_key = std::env::var("TENCENT_SECRET_KEY").ok();
243                    }
244                }
245                "voiceapi" => {
246                    // Set the endpoint from environment variable if not already set
247                    if self.endpoint.is_none() {
248                        self.endpoint = std::env::var("VOICEAPI_ENDPOINT")
249                            .ok()
250                            .or_else(|| Some("http://localhost:8000".to_string()));
251                    }
252                    // Set speaker ID from environment variable if not already set
253                    if self.speaker.is_none() {
254                        self.speaker = std::env::var("VOICEAPI_SPEAKER_ID")
255                            .ok()
256                            .or_else(|| Some("0".to_string()));
257                    }
258                }
259                "aliyun" => {
260                    if self.secret_key.is_none() {
261                        self.secret_key = std::env::var("DASHSCOPE_API_KEY").ok();
262                    }
263                }
264                "deepgram" => {
265                    if self.secret_key.is_none() {
266                        self.secret_key = std::env::var("DEEPGRAM_API_KEY").ok();
267                    }
268                }
269                _ => {}
270            }
271        }
272    }
273}
active_call/synthesis/mod.rs

active_call/synthesis/
mod.rs
