Spaces:
Running
Running
| import re | |
| import gradio as gr | |
| import numpy as np | |
| import tempfile | |
| from tqdm import tqdm | |
| from einops import rearrange | |
| from pydub import AudioSegment, silence | |
| from model import UNetT, DiT | |
| from cached_path import cached_path | |
| from model.utils import ( | |
| get_tokenizer, | |
| convert_char_to_pinyin, | |
| ) | |
| from infer.utils_infer import ( | |
| load_vocoder, | |
| load_model, | |
| remove_silence_edges, | |
| remove_silence_for_generated_wav, | |
| save_spectrogram, | |
| ) | |
| from tokenizers import Tokenizer | |
| from phonemizer import phonemize | |
| from transformers import pipeline | |
| import soundfile as sf | |
| # Import PyTorch and torchaudio | |
| import torch | |
| import torchaudio | |
| # Force CPU device for CPU Basic (free tier) | |
| device = torch.device("cpu") | |
| dtype = torch.float32 | |
| print(f"Using device: {device}, dtype: {dtype}") | |
| pipe = pipeline( | |
| "automatic-speech-recognition", | |
| model="openai/whisper-small", | |
| torch_dtype=dtype, | |
| device=device, | |
| ) | |
| vocos = load_vocoder() | |
| # --------------------- Settings -------------------- # | |
| target_sample_rate = 24000 | |
| n_mel_channels = 100 | |
| hop_length = 256 | |
| target_rms = 0.1 | |
| nfe_step = 16 | |
| cfg_strength = 2.0 | |
| ode_method = "euler" | |
| sway_sampling_coef = -1.0 | |
| speed = 1 | |
| fix_duration = None | |
| ref_language = "en-us" | |
| language = "en-us" | |
| DEFAULT_TTS_MODEL = "F5-TTS" | |
| tts_model_choice = DEFAULT_TTS_MODEL | |
| # load models | |
| F5TTS_model_cfg = dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4) | |
| E2TTS_model_cfg = dict(dim=1024, depth=24, heads=16, ff_mult=4) | |
| def load_custom(ckpt_path: str, vocab_path="", model_cfg=None): | |
| ckpt_path, vocab_path = ckpt_path.strip(), vocab_path.strip() | |
| if ckpt_path.startswith("hf://"): | |
| ckpt_path = str(cached_path(ckpt_path)) | |
| if vocab_path.startswith("hf://"): | |
| vocab_path = str(cached_path(vocab_path)) | |
| if model_cfg is None: | |
| model_cfg = dict(dim=1024, depth=22, heads=16, ff_mult=2, text_dim=512, conv_layers=4) | |
| return load_model(DiT, model_cfg, ckpt_path, vocab_file=vocab_path) | |
| custom_ema_model, pre_custom_path = None, "" | |
| chat_model_state = None | |
| chat_tokenizer_state = None | |
| # load models | |
| F5TTS_ema_model = load_custom( | |
| "hf://Gregniuki/F5-tts_English_German_Polish/multi3/model_900000.pt", "", F5TTS_model_cfg | |
| ) | |
| def chunk_text(text, max_chars): | |
| """ | |
| Splits the input text into chunks, ensuring: | |
| - Chunks are split by punctuation where possible. | |
| - If no punctuation is found and the chunk exceeds `split_after_space_chars`, | |
| it is split into smaller chunks of up to `split_after_space_chars`. | |
| Args: | |
| text (str): The text to be split. | |
| max_chars (int): The maximum number of characters per chunk after punctuation. | |
| split_after_space_chars (int): The maximum number of characters per chunk when no punctuation is present. | |
| Returns: | |
| List[str]: A list of text chunks. | |
| """ | |
| if max_chars > 135: | |
| max_chars = 135 | |
| if max_chars < 50: | |
| max_chars = 50 | |
| split_after_space_chars = max_chars + int(max_chars * 0.1) | |
| chunks = [] | |
| current_chunk = "" | |
| # Split the text into sentences based on punctuation followed by whitespace | |
| sentences = re.split(r"(?<=[;:,.。、!?])\s+|(?<=[；：,.。、！？])", text) | |
| for sentence in sentences: | |
| # If adding this sentence doesn't exceed max_chars, append it to the current chunk | |
| if len(current_chunk) + len(sentence) + 1 <= max_chars: # +1 for space | |
| current_chunk += sentence + " " | |
| else: | |
| # If current chunk exceeds split_after_space_chars, handle the splitting | |
| while len(current_chunk) > split_after_space_chars: | |
| split_index = current_chunk.rfind(" ", 0, split_after_space_chars) | |
| if split_index == -1: # No spaces to split; force split at 135 characters | |
| split_index = split_after_space_chars | |
| chunks.append(current_chunk[:split_index].strip()) | |
| current_chunk = current_chunk[split_index:].strip() | |
| # Add the current chunk to the list and start a new chunk | |
| if current_chunk: | |
| chunks.append(current_chunk.strip()) | |
| current_chunk = sentence + " " | |
| # If the remaining chunk exceeds split_after_space_chars, split it further | |
| while len(current_chunk) > split_after_space_chars: | |
| split_index = current_chunk.rfind(" ", 0, split_after_space_chars) | |
| if split_index == -1: # No spaces to split; force split at 135 characters | |
| split_index = split_after_space_chars | |
| chunks.append(current_chunk[:split_index].strip()) | |
| current_chunk = current_chunk[split_index:].strip() | |
| # Add any leftover chunk | |
| if current_chunk: | |
| chunks.append(current_chunk.strip()) | |
| return chunks | |
| def text_to_ipa(text, language=language): | |
| try: | |
| ipa_text = phonemize( | |
| text, | |
| language=language, | |
| backend='espeak', | |
| strip=False, | |
| preserve_punctuation=True, | |
| with_stress=True | |
| ) | |
| # Remove language markings like (en), (cmn), (de), (pl), (ru) | |
| ipa_text = re.sub(r'\([a-z]{2,3}\)', '', ipa_text) | |
| ipa_text = re.sub(r'tʃˈaɪniːzlˈe̞tə', '', ipa_text) | |
| ipa_text = re.sub(r'tʃˈaɪniːzɭˈetə', '', ipa_text) | |
| ipa_text = re.sub(r'dʒˈapəniːzlˈe̞tə', '', ipa_text) | |
| ipa_text = re.sub(r'dʒˈapəniːzɭˈetə', '', ipa_text) | |
| return ipa_text | |
| except Exception as e: | |
| print(f"Error processing text: {text}. Error: {e}") | |
| return None | |
| def infer_batch(ref_audio, ref_text, gen_text_batches, exp_name, remove_silence, cross_fade_duration=0.15, progress=gr.Progress()): | |
| if exp_name == "Multi": | |
| ema_model = F5TTS_ema_model | |
| audio, sr = ref_audio | |
| if audio.shape[0] > 1: | |
| audio = torch.mean(audio, dim=0, keepdim=True) | |
| rms = torch.sqrt(torch.mean(torch.square(audio))) | |
| if rms < target_rms: | |
| audio = audio * target_rms / rms | |
| if sr != target_sample_rate: | |
| resampler = torchaudio.transforms.Resample(sr, target_sample_rate) | |
| audio = resampler(audio) | |
| audio = audio.to(device) | |
| tokenizer = Tokenizer.from_file("data/Emilia_ZH_EN_pinyin/tokenizer.json") | |
| vocab_size = tokenizer.get_vocab_size() | |
| vocab = tokenizer.get_vocab() | |
| generated_waves = [] | |
| spectrograms = [] | |
| punctuation_weights = {",": 0, ".": 0, " ": 0} | |
| progress = tqdm(gen_text_batches) | |
| ipa_text_ref = text_to_ipa(ref_text, language=ref_language) | |
| print(ref_language) | |
| print(language) | |
| for i, gen_text in enumerate(progress): | |
| ipa_text_gen = text_to_ipa(gen_text, language=language) | |
| print(ipa_text_gen) | |
| text_list = ipa_text_ref + ipa_text_gen | |
| print(text_list) | |
| encoding = tokenizer.encode(text_list) | |
| tokens = encoding.tokens | |
| text_list = ' '.join(map(str, tokens)) | |
| final_text_list = [text_list] | |
| print(final_text_list) | |
| # Calculate reference audio length | |
| ref_audio_len = audio.shape[-1] // hop_length | |
| if fix_duration is not None: | |
| duration = int(fix_duration * target_sample_rate / hop_length) | |
| else: | |
| # Calculate text lengths with weights | |
| def calculate_weighted_length(text): | |
| length = len(text.encode("utf-8")) | |
| additional_length = sum(punctuation_weights.get(char, 0) for char in text) | |
| return length + additional_length | |
| ref_text_len = calculate_weighted_length(ref_text) | |
| gen_text_len = calculate_weighted_length(gen_text) | |
| duration = max(250, int(ref_audio_len) + int(((ref_audio_len / ref_text_len) * gen_text_len) / speed)) | |
| print(f"Chunk {i + 1}: Duration: {duration} speed {speed}") | |
| # inference | |
| with torch.inference_mode(): | |
| audio = audio.to(ema_model.device) | |
| final_text_list = [t.to(ema_model.device) if isinstance(t, torch.Tensor) else t for t in final_text_list] | |
| generated, _ = ema_model.sample( | |
| cond=audio, | |
| text=final_text_list, | |
| duration=duration, | |
| steps=nfe_step, | |
| cfg_strength=cfg_strength, | |
| sway_sampling_coef=sway_sampling_coef, | |
| ) | |
| # Process generated tensor | |
| generated = generated[:, ref_audio_len:, :] | |
| generated_mel_spec = rearrange(generated, "1 n d -> 1 d n") | |
| # Convert to float32 for vocos on CPU | |
| generated_mel_spec = generated_mel_spec.to(dtype=torch.float32) | |
| generated_wave = vocos.decode(generated_mel_spec) | |
| # Adjust wave RMS if needed | |
| if rms < target_rms: | |
| generated_wave = generated_wave * rms / target_rms | |
| # Convert to numpy | |
| generated_wave = generated_wave.squeeze().cpu().numpy() | |
| generated_waves.append(generated_wave) | |
| spectrograms.append(generated_mel_spec[0].cpu().numpy()) | |
| # Combine all generated waves with cross-fading | |
| if cross_fade_duration <= 0: | |
| final_wave = np.concatenate(generated_waves) | |
| else: | |
| final_wave = generated_waves[0] | |
| for i in range(1, len(generated_waves)): | |
| prev_wave = final_wave | |
| next_wave = generated_waves[i] | |
| cross_fade_samples = int(cross_fade_duration * target_sample_rate) | |
| cross_fade_samples = min(cross_fade_samples, len(prev_wave), len(next_wave)) | |
| if cross_fade_samples <= 0: | |
| final_wave = np.concatenate([prev_wave, next_wave]) | |
| continue | |
| prev_overlap = prev_wave[-cross_fade_samples:] | |
| next_overlap = next_wave[:cross_fade_samples] | |
| fade_out = np.linspace(1, 0, cross_fade_samples) | |
| fade_in = np.linspace(0, 1, cross_fade_samples) | |
| cross_faded_overlap = prev_overlap * fade_out + next_overlap * fade_in | |
| new_wave = np.concatenate([ | |
| prev_wave[:-cross_fade_samples], | |
| cross_faded_overlap, | |
| next_wave[cross_fade_samples:] | |
| ]) | |
| final_wave = new_wave | |
| # Remove silence | |
| if remove_silence: | |
| with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f: | |
| final_wave_float32 = final_wave.astype(np.float32) | |
| sf.write(f.name, final_wave_float32, target_sample_rate) | |
| aseg = AudioSegment.from_file(f.name) | |
| non_silent_segs = silence.split_on_silence(aseg, min_silence_len=1000, silence_thresh=-50, keep_silence=500) | |
| non_silent_wave = AudioSegment.silent(duration=0) | |
| for non_silent_seg in non_silent_segs: | |
| non_silent_wave += non_silent_seg | |
| aseg = non_silent_wave | |
| aseg.export(f.name, format="wav") | |
| final_wave_np, _ = sf.read(f.name) | |
| final_wave = torch.from_numpy(final_wave_np).float() | |
| if final_wave.dim() == 1: | |
| final_wave = final_wave.unsqueeze(0) | |
| else: | |
| final_wave = final_wave.T | |
| final_wave = final_wave.squeeze().cpu().numpy() | |
| # Create a combined spectrogram | |
| combined_spectrogram = np.concatenate(spectrograms, axis=1) | |
| with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp_spectrogram: | |
| spectrogram_path = tmp_spectrogram.name | |
| save_spectrogram(combined_spectrogram, spectrogram_path) | |
| return (target_sample_rate, final_wave), spectrogram_path | |
| def infer(ref_audio_orig, ref_text, gen_text, exp_name, remove_silence, cross_fade_duration=0.15): | |
| print(gen_text) | |
| gr.Info("Converting audio...") | |
| with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f: | |
| aseg = AudioSegment.from_file(ref_audio_orig) | |
| aseg = remove_silence_edges(aseg) + AudioSegment.silent(duration=150) | |
| non_silent_segs = silence.split_on_silence( | |
| aseg, min_silence_len=700, silence_thresh=-50, keep_silence=700 | |
| ) | |
| non_silent_wave = AudioSegment.silent(duration=0) | |
| for non_silent_seg in non_silent_segs: | |
| non_silent_wave += non_silent_seg | |
| aseg = non_silent_wave | |
| audio_duration = len(aseg) | |
| if audio_duration > 10000: | |
| gr.Warning("Audio is over 10s, clipping to only first 10s.") | |
| aseg = aseg[:10000] | |
| aseg.export(f.name, format="wav") | |
| ref_audio = f.name | |
| if not ref_text or not ref_text.strip(): | |
| gr.Info("No reference text provided, transcribing reference audio...") | |
| ref_text = pipe( | |
| ref_audio, | |
| chunk_length_s=15, | |
| batch_size=8, | |
| generate_kwargs={"task": "transcribe"}, | |
| return_timestamps=False, | |
| )["text"].strip() | |
| gr.Info("Finished transcription") | |
| else: | |
| gr.Info("Using custom reference text...") | |
| # Add the functionality to ensure it ends with ". " | |
| if not ref_text.endswith(". "): | |
| if ref_text.endswith("."): | |
| ref_text += " " | |
| else: | |
| ref_text += ". " | |
| audio_np, sr = sf.read(ref_audio) | |
| audio = torch.from_numpy(audio_np).float() | |
| if audio.dim() == 1: | |
| audio = audio.unsqueeze(0) | |
| else: | |
| audio = audio.T # soundfile: (frames, channels) -> torchaudio: (channels, frames) | |
| max_chars = int((len(ref_text.encode('utf-8')) / (audio.shape[-1] / sr) * (20 - audio.shape[-1] / sr ))) | |
| print(f"text: {max_chars} ") | |
| gen_text_batches = chunk_text(gen_text, max_chars=max_chars) | |
| print('ref_text', ref_text) | |
| for i, batch_text in enumerate(gen_text_batches): | |
| print(f'gen_text {i}', batch_text) | |
| gr.Info(f"Generating audio using {exp_name} in {len(gen_text_batches)} batches") | |
| return infer_batch((audio, sr), ref_text, gen_text_batches, exp_name, remove_silence, cross_fade_duration) | |
| def generate_podcast(script, speaker1_name, ref_audio1, ref_text1, speaker2_name, ref_audio2, ref_text2, exp_name, remove_silence): | |
| speaker_pattern = re.compile(f"^({re.escape(speaker1_name)}|{re.escape(speaker2_name)}):", re.MULTILINE) | |
| speaker_blocks = speaker_pattern.split(script)[1:] | |
| generated_audio_segments = [] | |
| for i in range(0, len(speaker_blocks), 2): | |
| speaker = speaker_blocks[i] | |
| text = speaker_blocks[i+1].strip() | |
| if speaker == speaker1_name: | |
| ref_audio = ref_audio1 | |
| ref_text = ref_text1 | |
| elif speaker == speaker2_name: | |
| ref_audio = ref_audio2 | |
| ref_text = ref_text2 | |
| else: | |
| continue | |
| audio, _ = infer(ref_audio, ref_text, text, exp_name, remove_silence) | |
| sr, audio_data = audio | |
| with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file: | |
| sf.write(temp_file.name, audio_data, sr) | |
| audio_segment = AudioSegment.from_wav(temp_file.name) | |
| generated_audio_segments.append(audio_segment) | |
| pause = AudioSegment.silent(duration=500) | |
| generated_audio_segments.append(pause) | |
| final_podcast = sum(generated_audio_segments) | |
| with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file: | |
| podcast_path = temp_file.name | |
| final_podcast.export(podcast_path, format="wav") | |
| return podcast_path | |
| def parse_speechtypes_text(gen_text): | |
| pattern = r'\((.*?)\)' | |
| tokens = re.split(pattern, gen_text) | |
| segments = [] | |
| current_emotion = 'Regular' | |
| for i in range(len(tokens)): | |
| if i % 2 == 0: | |
| text = tokens[i].strip() | |
| if text: | |
| segments.append({'emotion': current_emotion, 'text': text}) | |
| else: | |
| emotion = tokens[i].strip() | |
| current_emotion = emotion | |
| return segments | |
| def update_language(new_language): | |
| global language | |
| language = new_language | |
| def update_language1(new_ref_language): | |
| global ref_language | |
| ref_language = new_ref_language | |
| def update_speed(new_speed): | |
| global speed | |
| speed = new_speed | |
| return f"Speed set to: {speed}" | |
| with gr.Blocks() as app_credits: | |
| gr.Markdown(""" | |
| # Credits | |
| * [mrfakename](https://github.com/fakerybakery) for the original [online demo](https://huggingface.co/spaces/mrfakename/E2-F5-TTS) | |
| * [RootingInLoad](https://github.com/RootingInLoad) for the podcast generation | |
| * [jpgallegoar](https://github.com/jpgallegoar) for multiple speech-type generation | |
| """) | |
| with gr.Blocks() as app_tts: | |
| gr.Markdown("# Batched TTS") | |
| ref_audio_input = gr.Audio(label="Reference Audio", type="filepath") | |
| gen_text_input = gr.Textbox(label="Text to Generate", lines=10) | |
| model_choice = gr.Radio( | |
| choices=["Multi"], label="Choose TTS Model", value="Multi" | |
| ) | |
| gr.Markdown("#Select Reference Language") | |
| language_choice1 = gr.Dropdown( | |
| choices=["af", "nl", "en-us", "en-gb", "en-029", "en-gb-x-gbclan", "en-gb-x-rp", "en-gb-scotland", "en-gb-x-gbcwmd", "de", "lb", | |
| "my", "yue", "hak", "cmn", | |
| "an", "ca", "fr-be", "fr-fr", "fr-ch", "ht", "it", "pap", "pt-br", "pt", "ro", "es", "es-419", | |
| "as", "bn", "bpy", "gu", "hi", "kok", "mr", "ne", "or", "pa", "sd", "si", "ur", | |
| "am", "ar", "he", "mt", | |
| "be", "ru", "ru-lv", "uk", | |
| "ja", | |
| "ko", | |
| "id", "mi", "ms", | |
| "az", "ba", "cu", "kk", "ky", "nog", "tk", "tt", "tr", "ug", "uz", | |
| "vi-vn-x-central", "vi", "vi-vn-x-south", | |
| "kn", "ml", "ta", "te", | |
| "shn", "th", | |
| "cs", "pl", "sk", | |
| "da", "is", "nb", "sv", | |
| "fa", "fa-latn", "ku", | |
| "tn", "sw", | |
| "grc", "el", | |
| "et", "fi", "hu", "smj", | |
| "bs", "bg", "hr", "mk", "sr", "sl", | |
| "sq", "hy", "hyw", | |
| "om", | |
| "ka", | |
| "ga", "gd", "cy", | |
| "ltg", "lv", "lt", | |
| "gn", | |
| "quc", "qu", | |
| "nci", | |
| "kl", | |
| "chr", | |
| "haw", | |
| "la", | |
| "eo", "ia", "io", "lfn", "jbo", "py", "qdb", "qya", "piqd", "sjn" | |
| ], label="Choose Language", value="en-us" | |
| ) | |
| gr.Markdown("#Select Synthesized Language") | |
| language_choice = gr.Dropdown( | |
| choices=["af", "nl", "en-us", "en-gb", "en-029", "en-gb-x-gbclan", "en-gb-x-rp", "en-gb-scotland", "en-gb-x-gbcwmd", "de", "lb", | |
| "my", "yue", "hak", "cmn", | |
| "an", "ca", "fr-be", "fr-fr", "fr-ch", "ht", "it", "pap", "pt-br", "pt", "ro", "es", "es-419", | |
| "as", "bn", "bpy", "gu", "hi", "kok", "mr", "ne", "or", "pa", "sd", "si", "ur", | |
| "am", "ar", "he", "mt", | |
| "be", "ru", "ru-lv", "uk", | |
| "ja", | |
| "ko", | |
| "id", "mi", "ms", | |
| "az", "ba", "cu", "kk", "ky", "nog", "tk", "tt", "tr", "ug", "uz", | |
| "vi-vn-x-central", "vi", "vi-vn-x-south", | |
| "kn", "ml", "ta", "te", | |
| "shn", "th", | |
| "cs", "pl", "sk", | |
| "da", "is", "nb", "sv", | |
| "fa", "fa-latn", "ku", | |
| "tn", "sw", | |
| "grc", "el", | |
| "et", "fi", "hu", "smj", | |
| "bs", "bg", "hr", "mk", "sr", "sl", | |
| "sq", "hy", "hyw", | |
| "om", | |
| "ka", | |
| "ga", "gd", "cy", | |
| "ltg", "lv", "lt", | |
| "gn", | |
| "quc", "qu", | |
| "nci", | |
| "kl", | |
| "chr", | |
| "haw", | |
| "la", | |
| "eo", "ia", "io", "lfn", "jbo", "py", "qdb", "qya", "piqd", "sjn" | |
| ], label="Choose Language", value="en-us" | |
| ) | |
| generate_btn = gr.Button("Synthesize", variant="primary") | |
| with gr.Accordion("Advanced Settings", open=False): | |
| ref_text_input = gr.Textbox( | |
| label="Reference Text", | |
| info="Leave blank to automatically transcribe the reference audio. If you enter text it will override automatic transcription.", | |
| lines=2, | |
| ) | |
| remove_silence = gr.Checkbox( | |
| label="Remove Silences", | |
| info="The model tends to produce silences, especially on longer audio. We can manually remove silences if needed. Note that this is an experimental feature and may produce strange results. This will also increase generation time.", | |
| value=False, | |
| ) | |
| speed_slider = gr.Slider( | |
| label="Speed", | |
| minimum=0.3, | |
| maximum=2.0, | |
| value=1.0, | |
| step=0.1, | |
| info="Adjust the speed of the audio.", | |
| ) | |
| cross_fade_duration_slider = gr.Slider( | |
| label="Cross-Fade Duration (s)", | |
| minimum=0.0, | |
| maximum=1.0, | |
| value=0.15, | |
| step=0.01, | |
| info="Set the duration of the cross-fade between audio clips.", | |
| ) | |
| language_status = gr.Textbox(label="Current Language", interactive=False) | |
| ref_language_status = gr.Textbox(label="Reference Language", interactive=False) | |
| speed_slider.change(update_speed, inputs=speed_slider) | |
| language_choice.change(update_language, inputs=language_choice, outputs=language_status) | |
| language_choice1.change(update_language1, inputs=language_choice1, outputs=ref_language_status) | |
| audio_output = gr.Audio(label="Synthesized Audio") | |
| spectrogram_output = gr.Image(label="Spectrogram") | |
| generate_btn.click( | |
| infer, | |
| inputs=[ | |
| ref_audio_input, | |
| ref_text_input, | |
| gen_text_input, | |
| model_choice, | |
| remove_silence, | |
| cross_fade_duration_slider, | |
| ], | |
| outputs=[audio_output, spectrogram_output], | |
| ) | |
| def parse_emotional_text(gen_text): | |
| pattern = r'\((.*?)\)' | |
| tokens = re.split(pattern, gen_text) | |
| segments = [] | |
| current_emotion = 'Regular' | |
| for i in range(len(tokens)): | |
| if i % 2 == 0: | |
| text = tokens[i].strip() | |
| if text: | |
| segments.append({'emotion': current_emotion, 'text': text}) | |
| else: | |
| emotion = tokens[i].strip() | |
| current_emotion = emotion | |
| return segments | |
| with gr.Blocks() as app_emotional: | |
| gr.Markdown( | |
| """ | |
| # Multiple Speech-Type Generation | |
| This section allows you to upload different audio clips for each speech type. 'Regular' emotion is mandatory. You can add additional speech types by clicking the "Add Speech Type" button. Enter your text in the format shown below, and the system will generate speech using the appropriate emotions. If unspecified, the model will use the regular speech type. The current speech type will be used until the next speech type is specified. | |
| **Example Input:** | |
| (Regular) Hello, I'd like to order a sandwich please. (Surprised) What do you mean you're out of bread? (Sad) I really wanted a sandwich though... (Angry) You know what, darn you and your little shop, you suck! (Whisper) I'll just go back home and cry now. (Shouting) Why me?! | |
| """ | |
| ) | |
| gr.Markdown("Upload different audio clips for each speech type. 'Regular' emotion is mandatory. You can add additional speech types by clicking the 'Add Speech Type' button.") | |
| with gr.Row(): | |
| regular_name = gr.Textbox(value='Regular', label='Speech Type Name', interactive=False) | |
| regular_audio = gr.Audio(label='Regular Reference Audio', type='filepath') | |
| regular_ref_text = gr.Textbox(label='Reference Text (Regular)', lines=2) | |
| max_speech_types = 10 | |
| speech_type_names = [] | |
| speech_type_audios = [] | |
| speech_type_ref_texts = [] | |
| speech_type_delete_btns = [] | |
| for i in range(max_speech_types - 1): | |
| with gr.Row(): | |
| name_input = gr.Textbox(label='Speech Type Name', visible=False) | |
| audio_input = gr.Audio(label='Reference Audio', type='filepath', visible=False) | |
| ref_text_input = gr.Textbox(label='Reference Text', lines=2, visible=False) | |
| delete_btn = gr.Button("Delete", variant="secondary", visible=False) | |
| speech_type_names.append(name_input) | |
| speech_type_audios.append(audio_input) | |
| speech_type_ref_texts.append(ref_text_input) | |
| speech_type_delete_btns.append(delete_btn) | |
| add_speech_type_btn = gr.Button("Add Speech Type") | |
| speech_type_count = gr.State(value=0) | |
| def add_speech_type_fn(speech_type_count): | |
| if speech_type_count < max_speech_types - 1: | |
| speech_type_count += 1 | |
| name_updates = [] | |
| audio_updates = [] | |
| ref_text_updates = [] | |
| delete_btn_updates = [] | |
| for i in range(max_speech_types - 1): | |
| if i < speech_type_count: | |
| name_updates.append(gr.update(visible=True)) | |
| audio_updates.append(gr.update(visible=True)) | |
| ref_text_updates.append(gr.update(visible=True)) | |
| delete_btn_updates.append(gr.update(visible=True)) | |
| else: | |
| name_updates.append(gr.update()) | |
| audio_updates.append(gr.update()) | |
| ref_text_updates.append(gr.update()) | |
| delete_btn_updates.append(gr.update()) | |
| else: | |
| name_updates = [gr.update() for _ in range(max_speech_types - 1)] | |
| audio_updates = [gr.update() for _ in range(max_speech_types - 1)] | |
| ref_text_updates = [gr.update() for _ in range(max_speech_types - 1)] | |
| delete_btn_updates = [gr.update() for _ in range(max_speech_types - 1)] | |
| return [speech_type_count] + name_updates + audio_updates + ref_text_updates + delete_btn_updates | |
| add_speech_type_btn.click( | |
| add_speech_type_fn, | |
| inputs=speech_type_count, | |
| outputs=[speech_type_count] + speech_type_names + speech_type_audios + speech_type_ref_texts + speech_type_delete_btns | |
| ) | |
| def make_delete_speech_type_fn(index): | |
| def delete_speech_type_fn(speech_type_count): | |
| name_updates = [] | |
| audio_updates = [] | |
| ref_text_updates = [] | |
| delete_btn_updates = [] | |
| for i in range(max_speech_types - 1): | |
| if i == index: | |
| name_updates.append(gr.update(visible=False, value='')) | |
| audio_updates.append(gr.update(visible=False, value=None)) | |
| ref_text_updates.append(gr.update(visible=False, value='')) | |
| delete_btn_updates.append(gr.update(visible=False)) | |
| else: | |
| name_updates.append(gr.update()) | |
| audio_updates.append(gr.update()) | |
| ref_text_updates.append(gr.update()) | |
| delete_btn_updates.append(gr.update()) | |
| speech_type_count = max(0, speech_type_count - 1) | |
| return [speech_type_count] + name_updates + audio_updates + ref_text_updates + delete_btn_updates | |
| return delete_speech_type_fn | |
| for i, delete_btn in enumerate(speech_type_delete_btns): | |
| delete_fn = make_delete_speech_type_fn(i) | |
| delete_btn.click( | |
| delete_fn, | |
| inputs=speech_type_count, | |
| outputs=[speech_type_count] + speech_type_names + speech_type_audios + speech_type_ref_texts + speech_type_delete_btns | |
| ) | |
| gen_text_input_emotional = gr.Textbox(label="Text to Generate", lines=10) | |
| model_choice_emotional = gr.Radio( | |
| choices=["Multi"], label="Choose TTS Model", value="Multi" | |
| ) | |
| with gr.Accordion("Advanced Settings", open=False): | |
| remove_silence_emotional = gr.Radio( | |
| choices=["True", "False"], | |
| label="Remove Silences", | |
| value="False", | |
| info="Manually remove silences (experimental).", | |
| ) | |
| generate_emotional_btn = gr.Button("Generate Emotional Speech", variant="primary") | |
| audio_output_emotional = gr.Audio(label="Synthesized Audio") | |
| def generate_emotional_speech( | |
| regular_audio, | |
| regular_ref_text, | |
| gen_text, | |
| *args, | |
| ): | |
| num_additional_speech_types = max_speech_types - 1 | |
| speech_type_names_list = args[0:num_additional_speech_types] | |
| speech_type_audios_list = args[num_additional_speech_types : 2 * num_additional_speech_types] | |
| speech_type_ref_texts_list = args[2 * num_additional_speech_types : 3 * num_additional_speech_types] | |
| model_choice = args[3 * num_additional_speech_types] | |
| remove_silence_str = args[3 * num_additional_speech_types + 1] | |
| remove_silence = remove_silence_str == "True" | |
| speech_types = {'Regular': {'audio': regular_audio, 'ref_text': regular_ref_text}} | |
| for name_input, audio_input, ref_text_input in zip(speech_type_names_list, speech_type_audios_list, speech_type_ref_texts_list): | |
| if name_input and audio_input: | |
| speech_types[name_input] = {'audio': audio_input, 'ref_text': ref_text_input} | |
| segments = parse_speechtypes_text(gen_text) | |
| generated_audio_segments = [] | |
| current_emotion = 'Regular' | |
| for segment in segments: | |
| emotion = segment['emotion'] | |
| text = segment['text'] | |
| if emotion in speech_types: | |
| current_emotion = emotion | |
| else: | |
| current_emotion = 'Regular' | |
| ref_audio = speech_types[current_emotion]['audio'] | |
| ref_text = speech_types[current_emotion].get('ref_text', '') | |
| audio, _ = infer(ref_audio, ref_text, text, model_choice, remove_silence) | |
| sr, audio_data = audio | |
| generated_audio_segments.append(audio_data) | |
| if generated_audio_segments: | |
| final_audio_data = np.concatenate(generated_audio_segments) | |
| return (sr, final_audio_data) | |
| else: | |
| gr.Warning("No audio generated.") | |
| return None | |
| generate_emotional_btn.click( | |
| generate_emotional_speech, | |
| inputs=[ | |
| regular_audio, | |
| regular_ref_text, | |
| gen_text_input_emotional, | |
| ] + speech_type_names + speech_type_audios + speech_type_ref_texts + [ | |
| model_choice_emotional, | |
| remove_silence_emotional, | |
| ], | |
| outputs=audio_output_emotional, | |
| ) | |
| def validate_speech_types( | |
| gen_text, | |
| regular_name, | |
| *args | |
| ): | |
| num_additional_speech_types = max_speech_types - 1 | |
| speech_type_names_list = args[:num_additional_speech_types] | |
| speech_types_available = set() | |
| if regular_name: | |
| speech_types_available.add(regular_name) | |
| for name_input in speech_type_names_list: | |
| if name_input: | |
| speech_types_available.add(name_input) | |
| segments = parse_emotional_text(gen_text) | |
| speech_types_in_text = set(segment['emotion'] for segment in segments) | |
| missing_speech_types = speech_types_in_text - speech_types_available | |
| if missing_speech_types: | |
| return gr.update(interactive=False) | |
| else: | |
| return gr.update(interactive=True) | |
| gen_text_input_emotional.change( | |
| validate_speech_types, | |
| inputs=[gen_text_input_emotional, regular_name] + speech_type_names, | |
| outputs=generate_emotional_btn | |
| ) | |
| with gr.Blocks() as app: | |
| gr.Markdown( | |
| """ | |
| # F5 TTS (CPU Basic) | |
| This is a CPU-optimized web UI for F5 TTS with advanced batch processing support. This app supports the following TTS models: | |
| * [F5-TTS](https://arxiv.org/abs/2410.06885) (A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching) | |
| * [E2 TTS](https://arxiv.org/abs/2406.18009) (Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS) | |
| The checkpoint was trained with Polish British English American English German Russian Ukrainian other languages may not work correctly. | |
| If you're having issues, try converting your reference audio to WAV or MP3, clipping it to 10s, and shortening your prompt. | |
| **NOTE: This runs on CPU Basic. Generation will be slower than GPU. Reference text will be automatically transcribed with Whisper if not provided. For best results, keep your reference clips short (<15s). Ensure the audio is fully uploaded before generating.** | |
| """ | |
| ) | |
| gr.HTML( | |
| """ | |
| <a href="https://www.ko-fi.com/gregs40829" target="_blank"> | |
| <img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" style="height: 60px !important;width: 217px !important;"> | |
| </a> | |
| """ | |
| ) | |
| gr.TabbedInterface([app_tts, app_emotional, app_credits], ["TTS", "Multi-Style", "Credits"]) | |
| import click | |
| def main(port, host, share, api): | |
| global app | |
| print(f"Starting app on CPU...") | |
| app.queue(api_open=api).launch( | |
| server_name=host, server_port=port, share=share | |
| ) | |
| if __name__ == "__main__": | |
| main() | |