Update app.py

app.py

@@ -42,11 +42,12 @@ pipe_dict = {
     "language": "english",
 }
 
-title = "# VITS"
-
-description = """
-TODO
-"""
+title =      """# Explore English and Spanish Accents with VITS finetuning
+            ## Or how the best wine comes in old bottles
+            [VITS](https://huggingface.co/docs/transformers/model_doc/vits) is a light weight, low-latency TTS model. 
+            Coupled with the right datasets and the right training recipes, you can get an excellent finetuned version in 20 minutes with as little as 80 to 150 samples.            
+            Stay tuned, the training recipe is coming soon!
+            """,
 
 max_speakers = 15
 
@@ -80,12 +81,22 @@ def generate_audio(text, language):
     out.extend([gr.Audio(visible=False)]*(max_speakers-(len(out))))
     return out
 
+css = """
+#container{
+    margin: 0 auto;
+    max-width: 80rem;
+}
+#intro{
+    max-width: 100%;
+    text-align: center;
+    margin: 0 auto;
+}
+"""
 
 # Gradio blocks demo    
-with gr.Blocks() as demo_blocks:
+with gr.Blocks(css=css) as demo_blocks:
     gr.Markdown(title)
-    gr.Markdown(description)
-    with gr.Row():
+    with gr.Row(elem_id="container"):
         with gr.Column():
             inp_text = gr.Textbox(label="Input Text", info="What would you like VITS to synthesise?")
             btn = gr.Button("Generate Audio!")
@@ -101,6 +112,47 @@ with gr.Blocks() as demo_blocks:
             for i in range(max_speakers):
                 out_audio = gr.Audio(type="numpy", autoplay=False, label=f"Generated Audio - speaker {i}", show_label=True, visible=False)
                 outputs.append(out_audio)
+
+
+        gr.Markdown("""
+            ## Datasets and models details
+
+            ### English
+
+            * **Model**: [VITS-ljs](https://huggingface.co/kakao-enterprise/vits-ljs)
+            * **Dataset**: [British Isles Accent](https://huggingface.co/datasets/ylacombe/english_dialects). For each accent, we used 100 to 150 samples of a single speaker to finetune [VITS-ljs](https://huggingface.co/kakao-enterprise/vits-ljs).
+            
+            ### Spanish
+
+            * **Model**: [Spanish MMS TTS](https://huggingface.co/facebook/mms-tts-spa). This model is part of Facebook's [Massively Multilingual Speech](https://arxiv.org/abs/2305.13516) project, aiming to
+provide speech technology across a diverse range of languages. You can find more details about the supported languages
+and their ISO 639-3 codes in the [MMS Language Coverage Overview](https://dl.fbaipublicfiles.com/mms/misc/language_coverage_mms.html),
+and see all MMS-TTS checkpoints on the Hugging Face Hub: [facebook/mms-tts](https://huggingface.co/models?sort=trending&search=facebook%2Fmms-tts).
+            * **Datasets**:  For each accent, we used 100 to 150 samples of a single speaker to finetune the model.
+                - [Colombian Spanish TTS dataset](https://huggingface.co/datasets/ylacombe/google-colombian-spanish).
+                - [Argentinian Spanish TTS dataset](https://huggingface.co/datasets/ylacombe/google-argentinian-spanish).
+                - [Chilean Spanish TTS dataset](https://huggingface.co/datasets/ylacombe/google-chilean-spanish).
+            
+            """) 
+
+        with gr.Accordion("Run with transformers"):
+            gr.Markdown(
+                """## Running VITS and MMS with transformers
+            ```bash
+            pip install transformers
+            ```
+            ```py
+            from transformers import pipeline
+            import scipy
+            pipe = pipeline("text-to-speech", model="kakao-enterprise/vits-ljs", device=0)
+            
+            results = pipe("A cinematic shot of a baby racoon wearing an intricate italian priest robe")
+
+            # write to a wav file
+            scipy.io.wavfile.write("audio_vits.wav", rate=results["sampling_rate"], data=results["audio"].squeeze())
+            ```
+            """
+            )
     
     btn.click(generate_audio, [inp_text, language], outputs)