publish

app.py

@@ -232,8 +232,6 @@ DESCR = """
 # Japanese TTS Arena: Benchmarking Japanese TTS Models in the Wild
 
 Vote to help the community find the best available text-to-speech model!
-
-_This arena is inspired and built on [TTS Arena](https://huggingface.co/spaces/TTS-AGI/TTS-Arena)._
 """.strip()
 # INSTR = """
 # ## Instructions
@@ -264,7 +262,7 @@ Please [create a Discussion](https://huggingface.co/spaces/{SPACE_ID}/discussion
 ABOUT = f"""
 ## 📄 About
 
-The TTS Arena evaluates leading speech synthesis models. It is inspired by LMsys's [Chatbot Arena](https://chat.lmsys.org/).
+The Japanese TTS Arena evaluates leading speech synthesis models. It is inspired by LMsys's [Chatbot Arena](https://chat.lmsys.org/) and [TTS Arena](https://huggingface.co/spaces/TTS-AGI/TTS-Arena).
 
 ### Motivation
 
@@ -276,16 +274,7 @@ The leaderboard allows a user to enter text, which will be synthesized by two mo
 
 ### Credits
 
-Thank you to the following individuals who helped make this project possible:
-
-* VB ([Twitter](https://twitter.com/reach_vb) / [Hugging Face](https://huggingface.co/reach-vb))
-* Clémentine Fourrier ([Twitter](https://twitter.com/clefourrier) / [Hugging Face](https://huggingface.co/clefourrier))
-* Lucain Pouget ([Twitter](https://twitter.com/Wauplin) / [Hugging Face](https://huggingface.co/Wauplin))
-* Yoach Lacombe ([Twitter](https://twitter.com/yoachlacombe) / [Hugging Face](https://huggingface.co/ylacombe))
-* Main Horse ([Twitter](https://twitter.com/main_horse) / [Hugging Face](https://huggingface.co/main-horse))
-* Sanchit Gandhi ([Twitter](https://twitter.com/sanchitgandhi99) / [Hugging Face](https://huggingface.co/sanchit-gandhi))
-* Apolinário Passos ([Twitter](https://twitter.com/multimodalart) / [Hugging Face](https://huggingface.co/multimodalart))
-* Pedro Cuenca ([Twitter](https://twitter.com/pcuenq) / [Hugging Face](https://huggingface.co/pcuenq))
+Thank you to the open-source code from TTS Arena which helped make this project possible.
 
 {request}
 
@@ -296,13 +285,11 @@ We may store text you enter and generated audio. We store a unique ID for each s
 ### License
 
 Generated audio clips cannot be redistributed and may be used for personal, non-commercial use only.
-
-Random sentences are sourced from a filtered subset of the [Harvard Sentences](https://www.cs.columbia.edu/~hgs/audio/harvard.html).
 """.strip()
 LDESC = """
 ## 🏆 Leaderboard
 
-Vote to help the community determine the best text-to-speech (TTS) models.
+Vote to help the community determine the best Japanese text-to-speech (TTS) models.
 
 The leaderboard displays models in descending order of how natural they sound (based on votes cast by the community).
 
@@ -445,7 +432,8 @@ def get_leaderboard(reveal_prelim = False):
     cursor = conn.cursor()
     sql = 'SELECT name, upvote, downvote FROM model'
     # if not reveal_prelim: sql += ' WHERE EXISTS (SELECT 1 FROM model WHERE (upvote + downvote) > 750)'
-    if not reveal_prelim: sql += ' WHERE (upvote + downvote) > 500'
+    # if not reveal_prelim: sql += ' WHERE (upvote + downvote) > 500'
+    if not reveal_prelim: sql += ' WHERE (upvote + downvote) > 2'
     cursor.execute(sql)
     data = cursor.fetchall()
     df = pd.DataFrame(data, columns=['name', 'upvote', 'downvote'])
@@ -676,7 +664,7 @@ def synthandreturn(text):
     # Get two random models
     mdl1, mdl2 = random.sample(list(AVAILABLE_MODELS.keys()), 2)
     log_text(text)
-    print("[debug] Using", mdl1, mdl2)
+    # print("[debug] Using", mdl1, mdl2)
     def predict_and_update_result(text, model, result_storage):
         try:
             # if model in AVAILABLE_MODELS:
@@ -693,10 +681,6 @@ def synthandreturn(text):
                 # result = router.predict(text, AVAILABLE_MODELS[model].lower(), api_name="/synthesize")
                 if model in model_kwargs:
                     router = Client(model_links[model])
-                    # debug
-                    print(model_args[model])
-                    print(model_kwargs[model])
-
                     result = router.predict(*model_args[model], **model_kwargs[model])
                 else:
                     result = get_tts_file(text, model)
@@ -933,7 +917,7 @@ with gr.Blocks() as about:
 #         dbtext = gr.Textbox(label="Type \"delete db\" to confirm", placeholder="delete db")
 #         ddb = gr.Button("Delete DB")
 #     ddb.click(del_db, inputs=dbtext, outputs=ddb)
-with gr.Blocks(theme=theme, css="footer {visibility: hidden}textbox{resize:none}", title="TTS Arena copy") as demo:
+with gr.Blocks(theme=theme, css="footer {visibility: hidden}textbox{resize:none}", title="JA TTS Arena") as demo:
     gr.Markdown(DESCR)
     # gr.TabbedInterface([vote, leaderboard, about, admin], ['Vote', 'Leaderboard', 'About', 'Admin (ONLY IN BETA)'])
     gr.TabbedInterface([vote, leaderboard, about], ['🗳️ Vote', '🏆 Leaderboard', '📄 About'])