update

talk_arena/audio_collection.py

@@ -18,16 +18,21 @@ os.makedirs("outputs", exist_ok=True)
 # Initialize Hugging Face API client
 hf_api = HfApi(token=os.getenv("HF_TOKEN"))
 DATASET_REPO = "alisartazkhan/audioLLM_judge"
-CATEGORY = "pilot_tempo_control5"
+CATEGORY = "pilot_tempo_control6"
 MAX_RECORDINGS = 5  # Number of prompts to record
 COMPLETION_CODE = "CEO4RWQ6"
 resampler = Audio(sampling_rate=16_000)
 
-# Load the prompts from a JSON file
-prompt_path = os.path.join(os.path.dirname(__file__), "prompts.json")
-with open(prompt_path, "r") as f:
-    prompts_data = json.load(f)
-    PROMPTS = prompts_data["prompts"]
+# ====== MODIFY THIS SECTION TO CHANGE INSTRUCTIONS AND PROMPT ======
+# Instructions for the user
+USER_INSTRUCTIONS = """
+## Recording Instructions:
+Please record yourself reading your instruction clearly and naturally, speaking into the microphone in a quiet environment.
+"""
+
+# The prompt that users will record
+RECORDING_PROMPT = ""
+# ================================================================
 
 # Create a JSON database to track uploads
 class UploadTracker:
@@ -44,10 +49,10 @@ class UploadTracker:
             with open(filename, "r") as f:
                 self.data = json.load(f)
     
-    def add_recording(self, prompt_index, audio_hash, filename):
+    def add_recording(self, audio_hash, filename):
         """Add a record of an uploaded recording"""
         record = {
-            "prompt_index": prompt_index,
+            "prompt": RECORDING_PROMPT,
             "audio_hash": audio_hash,
             "filename": filename,
             "timestamp": str(uuid.uuid4())
@@ -98,17 +103,17 @@ def upload_to_hf(local_path, repo_path):
         print(f"Error uploading file to HF: {e}")
         return False
 
-def on_submit(audio_input, prompt_index):
+def on_submit(audio_input, recording_count):
     """Handle the submission of a recorded audio prompt"""
     if audio_input is None:
         return (
-            gr.Markdown(f"# Recording {prompt_index + 1}/{MAX_RECORDINGS}"),
-            gr.Markdown(f"## Please record yourself reading the following sentence:"),
-            gr.Markdown(f"### \"{PROMPTS[prompt_index]}\""),
+            gr.Markdown(f"# Recording {recording_count + 1}/{MAX_RECORDINGS}"),
+            gr.Markdown(USER_INSTRUCTIONS),
+            gr.Markdown(f"### \"{RECORDING_PROMPT}\""),
             gr.Audio(value=None, label="Record your response"),
             gr.Button("Submit Recording", interactive=False),
-            gr.Button("Next Prompt", visible=False),
-            prompt_index
+            gr.Button("Next Recording", visible=False),
+            recording_count
         )
     
     # Process the audio
@@ -123,67 +128,57 @@ def on_submit(audio_input, prompt_index):
 
     # Create unique filename
     unique_id = str(uuid.uuid4())[:8]
-    local_filename = f"outputs/prompt{prompt_index}_{audio_hash}_{unique_id}.wav"
+    clean_prompt = RECORDING_PROMPT.replace(" ", "_").replace(".", "").replace(",", "")[:20]
+    local_filename = f"outputs/{clean_prompt}_{audio_hash}_{unique_id}.wav"
 
     # Save the original (non-resampled) audio
     sf.write(local_filename, y, sr, format="wav")
 
     # Upload to HF dataset
-    hf_path = f"{CATEGORY}/prompt{prompt_index}_{audio_hash}_{unique_id}.wav"
+    hf_path = f"{CATEGORY}/{clean_prompt}_{audio_hash}_{unique_id}.wav"
     upload_to_hf(local_filename, hf_path)
 
     # Add to tracker
-    tracker.add_recording(prompt_index, audio_hash, hf_path)
+    tracker.add_recording(audio_hash, hf_path)
     
     # Show success message
     return (
-        gr.Markdown(f"# Recording {prompt_index + 1}/{MAX_RECORDINGS}"),
-        gr.Markdown(f"## Recording successfully uploaded!"),
-        gr.Markdown(f"### {PROMPTS[prompt_index]}"),
+        gr.Markdown(f"# Recording {recording_count + 1}/{MAX_RECORDINGS}"),
+        gr.Markdown(USER_INSTRUCTIONS),
+        gr.Markdown(f"### Recording successfully uploaded!"),
         gr.Audio(value=None, label="Record your response"),
         gr.Button("Submit Recording", interactive=False),
-        gr.Button("Next Prompt", visible=True),
-        prompt_index
+        gr.Button("Next Recording", visible=True),
+        recording_count
     )
 
-def next_prompt(prompt_index):
+def next_prompt(recording_count):
     """Move to the next prompt"""
-    prompt_index += 1
+    recording_count += 1
     
     # Check if we've gone through all prompts
-    if prompt_index >= min(len(PROMPTS), MAX_RECORDINGS):
+    if recording_count >= MAX_RECORDINGS:
         return (
             gr.Markdown(f"# All recordings complete! Completion code: {COMPLETION_CODE}"),
             gr.Markdown("## Thank you for your participation."),
-            gr.Markdown("### You have completed all prompts."),
+            gr.Markdown("### You have completed all recordings."),
             gr.Audio(visible=False),
             gr.Button(visible=False),
             gr.Button(visible=False),
-            prompt_index
+            recording_count
         )
     
-    # Display the next prompt
+    # Display the next recording screen
     return (
-        gr.Markdown(f"# Recording {prompt_index + 1}/{MAX_RECORDINGS}"),
-        gr.Markdown(f"##  Please record yourself reading the following sentence:"),
-        gr.Markdown(f"### \"{PROMPTS[prompt_index]}\""),
+        gr.Markdown(f"# Recording {recording_count + 1}/{MAX_RECORDINGS}"),
+        gr.Markdown(USER_INSTRUCTIONS),
+        gr.Markdown(f"### \"{RECORDING_PROMPT}\""),
         gr.Audio(value=None, label="Record your response", sources=["microphone"]),
         gr.Button("Submit Recording", interactive=False),
-        gr.Button("Next Prompt", visible=False),
-        prompt_index
+        gr.Button("Next Recording", visible=False),
+        recording_count
     )
 
-def submit_and_next(audio_input, prompt_index):
-    """Handle submission and move to next prompt immediately"""
-    if audio_input is None:
-        return on_submit(audio_input, prompt_index)
-
-    # Call submission logic
-    _ = on_submit(audio_input, prompt_index)
-    
-    # Then move to next prompt
-    return next_prompt(prompt_index)
-
 def enable_submit_button(audio_input):
     """Enable the submit button when audio is recorded"""
     if audio_input is not None:
@@ -199,11 +194,11 @@ theme = gr.themes.Soft(
 
 # Create Gradio interface
 with gr.Blocks(theme=theme, css="footer {visibility: hidden}") as demo:
-    prompt_index = gr.State(0)
+    recording_count = gr.State(0)
     
     title = gr.Markdown(f"# Recording 1/{MAX_RECORDINGS}")
-    instructions = gr.Markdown("## Please record yourself reading the following sentence:")
-    prompt_text = gr.Markdown(f"### \"{PROMPTS[0]}\"")
+    instructions = gr.Markdown(USER_INSTRUCTIONS)
+    prompt_text = gr.Markdown(f"### \"{RECORDING_PROMPT}\"")
     
     audio_input = gr.Audio(
         label="Record your response", 
@@ -213,7 +208,7 @@ with gr.Blocks(theme=theme, css="footer {visibility: hidden}") as demo:
     
     with gr.Row():
         submit_btn = gr.Button("Submit Recording", interactive=False)
-        next_btn = gr.Button("Next Prompt", visible=False)
+        next_btn = gr.Button("Next Recording", visible=False)
     
     # Enable submit button when audio is recorded
     audio_input.change(
@@ -224,25 +219,18 @@ with gr.Blocks(theme=theme, css="footer {visibility: hidden}") as demo:
     
     # Handle submission
     submit_btn.click(
-        fn=submit_and_next,
-        inputs=[audio_input, prompt_index],
-        outputs=[title, instructions, prompt_text, audio_input, submit_btn, next_btn, prompt_index]
+        fn=on_submit,
+        inputs=[audio_input, recording_count],
+        outputs=[title, instructions, prompt_text, audio_input, submit_btn, next_btn, recording_count]
     )
-
     
     # Handle next button
     next_btn.click(
         fn=next_prompt,
-        inputs=[prompt_index],
-        outputs=[title, instructions, prompt_text, audio_input, submit_btn, next_btn, prompt_index]
+        inputs=[recording_count],
+        outputs=[title, instructions, prompt_text, audio_input, submit_btn, next_btn, recording_count]
     )
 
 # Launch the app
 if __name__ == "__main__":
-    # First, create the prompts.json file
-    with open("talkarena/prompts.json", "w") as f:
-        json.dump({
-            "prompts": PROMPTS
-        }, f, indent=2)
-    
     demo.launch(share=True)