diff --git a/crates/service/job/inference_job/src/job/job_types/f5_tts/f5_tts_inference_command.rs b/crates/service/job/inference_job/src/job/job_types/f5_tts/f5_tts_inference_command.rs index 450f736b2a..9359cca753 100644 --- a/crates/service/job/inference_job/src/job/job_types/f5_tts/f5_tts_inference_command.rs +++ b/crates/service/job/inference_job/src/job/job_types/f5_tts/f5_tts_inference_command.rs @@ -52,14 +52,9 @@ pub struct InferenceArgs<'s> { pub reference_audio_path: &'s Path, pub reference_transcript_path: Option<&'s Path>, - pub reference_language : Option<&'s str>, - //pub output_audio_directory: &'s Path, - pub output_file_path: &'s Path, - - pub maybe_reference_free: Option, - pub maybe_temperature: Option, - pub maybe_target_language: Option<&'s str>, + pub output_audio_directory: &'s Path, + // pub output_file_path: &'s Path, } impl F5TTSInferenceCommand { @@ -153,24 +148,12 @@ impl F5TTSInferenceCommand { } } - command.push_str(&format!(" --target_text {}", path_to_string(args.input_text_file))); + command.push_str(&format!(" --gen_file {}", path_to_string(args.input_text_file))); command.push_str(&format!(" --ref_audio {}", path_to_string(args.reference_audio_path))); //command.push_str(&format!(" --output_path {}", path_to_string(args.output_audio_directory))); - command.push_str(&format!(" --output_path {}", path_to_string(args.output_file_path))); - - if let Some(reference_free) = args.maybe_reference_free { - command.push_str(&format!(" --ref_free {}", reference_free)); - } - - if let Some(temperature) = args.maybe_temperature { - command.push_str(&format!(" --temperature {}", temperature)); - } - - if let Some(target_language) = args.maybe_target_language { - command.push_str(&format!(" --target_language {}", target_language)); - } + command.push_str(&format!(" --output_dir {}", path_to_string(args.output_audio_directory))); if let Some(maybe_docker_options) = self.maybe_docker_options.as_ref() { command = maybe_docker_options.to_command_string(&command); diff --git a/crates/service/job/inference_job/src/job/job_types/f5_tts/tts_inference/process_single_f5_tts_job.rs b/crates/service/job/inference_job/src/job/job_types/f5_tts/tts_inference/process_single_f5_tts_job.rs index ffde147b2f..22f64e32d1 100644 --- a/crates/service/job/inference_job/src/job/job_types/f5_tts/tts_inference/process_single_f5_tts_job.rs +++ b/crates/service/job/inference_job/src/job/job_types/f5_tts/tts_inference/process_single_f5_tts_job.rs @@ -68,7 +68,7 @@ pub async fn process_single_f5_tts_inference_job( .map_err(|e| ProcessSingleJobError::from_io_error(e))?; let output_dir = work_temp_dir.path().join("output"); - let output_file_path = output_dir.join("GPTSoVITS_1.wav"); + let output_file_path = output_dir.join("out.wav"); if !output_dir.exists() { std::fs::create_dir_all(&output_dir) @@ -164,7 +164,6 @@ pub async fn process_single_f5_tts_inference_job( original_media_upload_fs_path }; - const DEFAULT_LANGUAGE : &str = "english"; let inference_start_time = Instant::now(); @@ -176,11 +175,7 @@ pub async fn process_single_f5_tts_inference_job( reference_audio_path: &original_media_upload_fs_path, input_text_file: &text_input_fs_path, reference_transcript_path: None, - reference_language: Some(DEFAULT_LANGUAGE), - output_file_path: &output_file_path, - maybe_reference_free: None, - maybe_temperature: None, - maybe_target_language: Some(DEFAULT_LANGUAGE), + output_audio_directory: &output_dir, }); let inference_duration = Instant::now().duration_since(inference_start_time); diff --git a/crates/service/web/storyteller_web/src/http_server/routes/add_routes.rs b/crates/service/web/storyteller_web/src/http_server/routes/add_routes.rs index 7f0b0b498c..e1f9800d5b 100644 --- a/crates/service/web/storyteller_web/src/http_server/routes/add_routes.rs +++ b/crates/service/web/storyteller_web/src/http_server/routes/add_routes.rs @@ -300,6 +300,10 @@ fn add_tts_routes (app: App) -> App app.service(web::resource("/v1/tts/inference") .route(web::post().to(enqueue_infer_tts_handler)) .route(web::head().to(|| HttpResponse::Ok())) + ) + .service(web::resource("/v1/tts/f5_inference") + .route(web::post().to(enqueue_infer_f5_tts_handler)) + .route(web::head().to(|| HttpResponse::Ok())) ) // NB(bt,2024-04-03): Older root /tts/* routes should be drained of traffic. They are used in the Twitch Streamer API though. .service(