update audio split

app.py

@@ -46,42 +46,32 @@ vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(devic
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0).to(device)
 
-def split_text_into_chunks(text, max_length):
-    # Split text into chunks that fit within the maximum token limit
-    words = text.split()  # Split the text into words
-    chunks = []
-    current_chunk = []
-
-    for word in words:
-        if len(current_chunk) + len(word) + 1 <= max_length:  # +1 accounts for space between words
-            current_chunk.append(word)
-        else:
-            # Join the current chunk into a string and append it to the list of chunks
-            chunks.append(' '.join(current_chunk))
-            # Start a new chunk
-            current_chunk = [word]
-
-    # Append the last chunk
-    if current_chunk:
-        chunks.append(' '.join(current_chunk))
-
-    return chunks
+def split_text_into_chunks(text, processor, max_length):
+    # Tokenize the input text
+    tokens = processor(text, return_tensors="pt", truncation=False).input_ids[0]
+    
+    # Split tokens into chunks
+    chunked_tokens = [tokens[i:i + max_length] for i in range(0, len(tokens), max_length)]
+    
+    return chunked_tokens
 
 def textToSpeech(expression):
-    # Split the input text into smaller chunks if necessary
-    max_token_limit = 600  # Set the token limit for your model
-    chunks = split_text_into_chunks(expression, max_token_limit)
+    # Set the maximum token limit for your model
+    max_token_limit = 600  # The maximum token length the model can handle
+    
+    # Tokenize and split the text into chunks based on tokens
+    token_chunks = split_text_into_chunks(expression, processor, max_token_limit)
     
     # Initialize an empty list to store the generated speech chunks
     audio_chunks = []
 
-    # Process each chunk through the model
-    for chunk in chunks:
-        inputs = processor(text=chunk, return_tensors="pt").to(device)
-        speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
+    # Process each token chunk through the model
+    for token_chunk in token_chunks:
+        # Generate speech from the token chunk
+        speech = model.generate_speech(token_chunk.unsqueeze(0), speaker_embeddings, vocoder=vocoder)
         audio_chunks.append(speech.cpu().numpy())
 
-    # Concatenate the audio chunks into a single file
+    # Concatenate the audio chunks into a single array
     final_audio = np.concatenate(audio_chunks)
 
     # Get the current date and time for unique file naming

hell 3.10.15

@@ -0,0 +1,95 @@
+commit b8c7a08b2bbcfdee152786b6a7212a047c50fc85 (HEAD -> main, origin/main, origin/HEAD)
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 14:21:55 2024 +0200
+
+    state
+
+commit 469ea347d5b8ea6d5404d7e88c4b4ab3f4abea07
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 14:15:30 2024 +0200
+
+    update history
+
+commit 3665a4d6ca95c3a000e58ce1cb8a96fd34ecd9ce
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 14:11:30 2024 +0200
+
+    print
+
+commit 96ad9b6e84215b238a689f0a5e4d33971114a2d7
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 14:08:38 2024 +0200
+
+    update history return
+
+commit 46e5c3d007799e62d711648538998e777fabca81
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 14:02:18 2024 +0200
+
+    delete history
+
+commit a997e436dc12c9d1c0824216b9c274ffee7371f8
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:53:47 2024 +0200
+
+    update history
+
+commit a743d0417f8168714ec629191b0d0c228c5b236e
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:51:41 2024 +0200
+
+    update
+
+commit 2c12fc659c9e0f58f725b751b244de305ef31423
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:48:20 2024 +0200
+
+    update
+
+commit c1aaa3ef2578af41e237a507c82ba8810b23129f
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:45:29 2024 +0200
+
+    update new_history
+
+commit cbc1d5399d3c2b2ea2385503a96194c2aea7e5c7
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:41:12 2024 +0200
+
+    update
+
+commit 5605340f569f7de28844f49d7c8216159326240a
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Oct 6 13:26:52 2024 +0200
+
+    update 20241006
+
+commit 9819d723446298f3846c4eca55c081d1ab6083a4
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Mon Sep 30 16:22:08 2024 +0200
+
+    update markdown
+
+commit 60c04ae96546d672b6804699cd99623e15dea0b1
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Mon Sep 30 15:24:05 2024 +0200
+
+    call print
+
+commit 08681846ade0af0fa2323113c1416fc0c2a5e345
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Mon Sep 30 15:06:04 2024 +0200
+
+    update
+
+commit d99be9e3ac9f8e8148f3da866f94ae539e29d363
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Mon Sep 30 14:47:27 2024 +0200
+
+    edit prompt
+
+commit 23136538c718c9a7683a5f9f729790050986066b (grafted)
+Author: LichtLiu <LichtLiu@users.noreply.huggingface.co>
+Date:   Sun Sep 29 19:22:47 2024 +0200
+
+    update