213

README.md

@@ -4,7 +4,7 @@ emoji: 🎺
 colorFrom: blue
 colorTo: pink
 sdk: gradio
-sdk_version: 6.14.0
+sdk_version: 5.15.0
 app_file: app.py
 pinned: true
 short_description: Get a music sample inspired by the mood of an image

app.py

@@ -19,7 +19,7 @@ def check_api(model_name):
             return "api not ready yet"
     elif model_name == "AudioLDM-2":
         try :
-            client = Client("fffiloni/audioldm2-text2audio-text2music-API")
+            client = Client("fffiloni/audioldm2-text2audio-text2music-API", hf_token=hf_token)
             return "api ready"
         except : 
             return "api not ready yet"
@@ -31,7 +31,7 @@ def check_api(model_name):
             return "api not ready yet"
     elif model_name == "Mustango":
         try :
-            client = Client("fffiloni/mustango-API-2")
+            client = Client("fffiloni/mustango-API", hf_token=hf_token)
             return "api ready"
         except : 
             return "api not ready yet"
@@ -47,12 +47,6 @@ def check_api(model_name):
             return "api ready"
         except:
             return "api not ready yet"
-    elif model_name == "ACE Step":
-        try :
-            client = Client("fffiloni/ACE-Step-API", hf_token=hf_token)
-            return "api ready"
-        except : 
-            return "api not ready yet"
     
         
 from moviepy.editor import VideoFileClip
@@ -75,7 +69,7 @@ def extract_audio(video_in):
 
 
 def get_caption(image_in):
-    kosmos2_client = Client("fffiloni/Kosmos-2-API")
+    kosmos2_client = Client("fffiloni/Kosmos-2-API", hf_token=hf_token)
     kosmos2_result = kosmos2_client.predict(
 		image_input=handle_file(image_in),
 		text_input="Detailed",
@@ -131,10 +125,10 @@ def get_magnet(prompt):
     return result[1]
 
 def get_audioldm(prompt):
-    client = Client("fffiloni/audioldm2-text2audio-text2music-API")
+    client = Client("fffiloni/audioldm2-text2audio-text2music-API", hf_token=hf_token)
     seed = random.randint(0, MAX_SEED)
     result = client.predict(
-        prompt=prompt,	# str in 'Input text' Textbox component
+        text=prompt,	# str in 'Input text' Textbox component
         negative_prompt="Low quality.",	# str in 'Negative prompt' Textbox component
         duration=10,	# int | float (numeric value between 5 and 15) in 'Duration (seconds)' Slider component
         guidance_scale=6.5,	# int | float (numeric value between 0 and 7) in 'Guidance scale' Slider component
@@ -159,7 +153,7 @@ def get_riffusion(prompt):
     return result[1]
 
 def get_mustango(prompt):
-    client = Client("fffiloni/mustango-API-2")
+    client = Client("fffiloni/mustango-API", hf_token=hf_token)
     result = client.predict(
 		prompt=prompt,	# str  in 'Prompt' Textbox component
 		steps=200,	# float (numeric value between 100 and 200) in 'Steps' Slider component
@@ -191,39 +185,6 @@ def get_stable_audio_open(prompt):
     print(result)
     return result
 
-def get_ace(prompt):
-    from gradio_client import Client, handle_file
-
-    client = Client("fffiloni/ACE-Step-API", hf_token=hf_token)
-    result = client.predict(
-		audio_duration=-1,
-		prompt=prompt,
-		lyrics="[inst]",
-		infer_step=60,
-		guidance_scale=15,
-		scheduler_type="euler",
-		cfg_type="apg",
-		omega_scale=10,
-		manual_seeds=None,
-		guidance_interval=0.5,
-		guidance_interval_decay=0,
-		min_guidance_scale=3,
-		use_erg_tag=True,
-		use_erg_lyric=False,
-		use_erg_diffusion=True,
-		oss_steps=None,
-		guidance_scale_text=0,
-		guidance_scale_lyric=0,
-		audio2audio_enable=False,
-		ref_audio_strength=0.5,
-		ref_audio_input=None,
-		lora_name_or_path="none",
-		api_name="/__call__"
-    )
-    print(result)
-    return result[0]
-
-
 import re
 import torch
 from transformers import pipeline
@@ -251,7 +212,7 @@ Immediately STOP after that. It should be EXACTLY in this format:
 "The song is an instrumental. The song is in medium tempo with a classical guitar playing a lilting melody in accompaniment style. The song is emotional and romantic. The song is a romantic instrumental song. The chord sequence is Gm, F6, Ebm. The time signature is 4/4. This song is in Adagio. The key of this song is G minor."
 """
 
-@spaces.GPU()
+@spaces.GPU(enable_queue=True)
 def get_musical_prompt(user_prompt, chosen_model):
 
     """
@@ -276,34 +237,13 @@ def get_musical_prompt(user_prompt, chosen_model):
     print(f"SUGGESTED Musical prompt: {cleaned_text}")
     return cleaned_text.lstrip("\n")
 
-def infer(image_in, chosen_model):
-    """
-    Generate music from an input image and selected music generation model.
-
-    This function performs the following steps:
-    1. Checks that an image and a model have been provided.
-    2. Verifies if the selected model's API is currently available.
-    3. Uses an image captioning model (Kosmos-2) to describe the image.
-    4. Generates a musical prompt from the image caption using a language model.
-    5. Sends the musical prompt to the selected music generation model and retrieves the result.
-
-    Args:
-        image_in: The filepath to an input image. This image is used as inspiration to generate music.
-        chosen_model: The name of the model to use for music generation. Supported values include: "Mustango", "AudioLDM-2", "Riffusion", "ACE Step", "Stable Audio Open".
-
-    Returns:
-        - A string containing the musical prompt generated from the image.
-        - A flag to show the retry button in the UI (for user to edit and retry the generation).
-        - The output of the selected model, typically an audio filepath or object depending on model.
-    """
+def infer(image_in, chosen_model, api_status):
     if image_in == None :
         raise gr.Error("Please provide an image input")
 
     if chosen_model == [] :
         raise gr.Error("Please pick a model")
 
-    api_status = check_api(chosen_model)
-
     if api_status == "api not ready yet" :
         raise gr.Error("This model is not ready yet, you can pick another one instead :)")
     
@@ -332,9 +272,6 @@ def infer(image_in, chosen_model):
     elif chosen_model == "Stable Audio Open" :
         gr.Info("Now calling Stable Audio Open for music...")
         music_o = get_stable_audio_open(musical_prompt)
-    elif chosen_model == "ACE Step" :
-        gr.Info("Now calling ACE Step for music...")
-        music_o = get_ace(musical_prompt)
     
     return gr.update(value=musical_prompt, interactive=True), gr.update(visible=True), music_o
 
@@ -360,9 +297,6 @@ def retry(chosen_model, caption):
     elif chosen_model == "Stable Audio Open" :
         gr.Info("Now calling Stable Audio Open for music...")
         music_o = get_stable_audio_open(musical_prompt)
-    elif chosen_model == "ACE Step" :
-        gr.Info("Now calling ACE Step for music...")
-        music_o = get_ace(musical_prompt)
 
     return music_o
 
@@ -407,12 +341,11 @@ with gr.Blocks(css=css) as demo:
                         label = "Choose a model",
                         choices = [
                             #"MAGNet",
-                            #"ACE Step",
                             "AudioLDM-2",
                             "Riffusion",
                             "Mustango",
                             #"MusicGen",
-                            #"Stable Audio Open"
+                            "Stable Audio Open"
                         ],
                         value = None,
                         filterable = False
@@ -461,15 +394,13 @@ with gr.Blocks(css=css) as demo:
         fn = check_api,
         inputs = chosen_model,
         outputs = check_status,
-        queue = False,
-        api_visibility='undocumented'
+        queue = False
     )
 
     retry_btn.click(
         fn = retry,
         inputs = [chosen_model, caption],
-        outputs = [result],
-        api_visibility='undocumented'
+        outputs = [result]
     )
     
     submit_btn.click(
@@ -477,7 +408,7 @@ with gr.Blocks(css=css) as demo:
         inputs = [
             image_in,
             chosen_model,
-            #check_status
+            check_status
         ],
         outputs =[
             caption,
@@ -486,4 +417,4 @@ with gr.Blocks(css=css) as demo:
         ]
     )
 
-demo.queue(max_size=16).launch(show_error=True, ssr_mode=False, mcp_server=True)
+demo.queue(max_size=16).launch(show_api=False, show_error=True)