High-Quality Voice Cloning TTS for 600+ Languages
Python
fix: instruct in infer_batch
Pol Pastells19f6d7b99bd3
diff --git a/omnivoice/cli/infer_batch.py b/omnivoice/cli/infer_batch.pyindex f14b06d..82ad1af 100644--- a/omnivoice/cli/infer_batch.py+++ b/omnivoice/cli/infer_batch.py@@ -278,7 +278,7 @@ def cluster_samples_by_duration() -> List[List[Tuple]]:sample_with_duration = []for sample in samples:- save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd = sample+ save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd, instruct = sampletotal_duration = estimate_sample_total_duration(duration_estimator,text,@@ -321,7 +321,7 @@ def cluster_samples_by_batch_size("""Split samples into fixed-size batches, sorted by duration to minimize padding."""sample_with_duration = []for sample in samples:- save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd = sample+ save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd, instruct = sampletotal_duration = estimate_sample_total_duration(duration_estimator,text,@@ -359,9 +359,10 @@ def run_inference_batch(langs = []durations = []speeds = []+ instructs = []for sample in batch_samples:- save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd = sample+ save_name, ref_text, ref_audio_path, text, lang_id, lang_name, dur, spd, instruct = samplesave_names.append(save_name)ref_texts.append(ref_text)ref_audio_paths.append(ref_audio_path)@@ -369,6 +370,7 @@ def run_inference_batch(langs.append(lang_id)durations.append(dur)speeds.append(spd)+ instructs.append(instruct)start_time = time.time()audios = worker_model.generate(@@ -378,6 +380,7 @@ def run_inference_batch(ref_text=ref_texts,duration=durations if any(d is not None for d in durations) else None,speed=speeds if any(s is not None for s in speeds) else None,+ instruct=instructs if any(i is not None for i in instructs) else None,**gen_kwargs,)batch_synth_time = time.time() - start_time@@ -443,6 +446,7 @@ def main():lang_name,s.get("duration"),s.get("speed"),+ s.get("instruct"),))diff --git a/omnivoice/utils/data_utils.py b/omnivoice/utils/data_utils.pyindex 0618cf0..075be64 100644--- a/omnivoice/utils/data_utils.py+++ b/omnivoice/utils/data_utils.py@@ -58,6 +58,7 @@ def read_test_list(path):"language_name": obj.get("language_name"),"duration": obj.get("duration"),"speed": obj.get("speed"),+ "instruct": obj.get("instruct"),}samples.append(sample)return samples