upgrade to gradio 5.5.0, add Resemble Enhance (#420)

* upgrade to gradio 5.5.0

* add Resemble Enhance

.gitignore

@@ -34,6 +34,7 @@ config.json
 /data/models/audiocraft_plus/
 /data/models/parler_tts/
 /data/models/musicgen_audiogen/*/
+/data/models/resemble_enhance/ds/
 
 # Ignore temporary files
 temp/

data/models/resemble_enhance/hparams.yaml

@@ -0,0 +1,38 @@
+fg_dir: !!python/object/apply:pathlib.Path
+- data
+- fg
+bg_dir: !!python/object/apply:pathlib.Path
+- data
+- bg
+rir_dir: !!python/object/apply:pathlib.Path
+- data
+- rir
+load_fg_only: false
+wav_rate: 44100
+n_fft: 2048
+win_size: 2048
+hop_size: 420
+num_mels: 128
+stft_magnitude_min: 0.0001
+preemphasis: 0.97
+mix_alpha_range:
+- 0.2
+- 0.8
+nj: 64
+training_seconds: 3.0
+batch_size_per_gpu: 32
+min_lr: 1.0e-05
+max_lr: 0.0001
+warmup_steps: 1000
+max_steps: 1000000
+gradient_clipping: 1.0
+cfm_solver_method: midpoint
+cfm_solver_nfe: 64
+cfm_time_mapping_divisor: 4
+univnet_nc: 96
+lcfm_latent_dim: 64
+lcfm_training_mode: cfm
+lcfm_z_scale: 6
+vocoder_extra_dim: 32
+gan_training_start_step: null
+praat_augment_prob: 0.2

extensions.json

@@ -254,6 +254,21 @@
             "website": "https://github.com/rsxdalv/extension_model_downloader",
             "extension_website": "https://github.com/rsxdalv/extension_model_downloader",
             "extension_platform_version": "0.0.1"
+        },
+        {
+            "package_name": "extension_resemble_enhance",
+            "name": "Resemble Enhance",
+            "version": "0.0.1",
+            "requirements": "git+https://github.com/rsxdalv/extension_resemble_enhance@main",
+            "description": "Resemble Enhance allows enhancing audio files.",
+            "extension_type": "interface",
+            "extension_class": "audio-conversion",
+            "author": "rsxdalv",
+            "extension_author": "rsxdalv",
+            "license": "MIT",
+            "website": "https://github.com/rsxdalv/extension_resemble_enhance",
+            "extension_website": "https://github.com/rsxdalv/extension_resemble_enhance",
+            "extension_platform_version": "0.0.1"
         }
     ],
     "decorators": [

installer_scripts/versions.json

@@ -1,6 +1,6 @@
 {
     "version": "0.2.0",
-    "pip_packages": 8,
+    "pip_packages": 9,
     "npm_packages": 4,
     "react_ui": 4
 }

requirements.txt

@@ -1,6 +1,6 @@
 safetensors
 ffmpeg-python # Apache 2.0
-gradio==5.1.0 # MIT License
+gradio==5.5.0 # MIT License
 python-dotenv==1.0.0
 soundfile==0.12.1; sys_platform == 'win32' # torchaudio
 # sox # torchaudio for linux

tts_webui/extensions_loader/interface_extensions.py

@@ -46,8 +46,8 @@ def _handle_package(package_name, title_name, requirements):
                         update_button()
                     else:
                         _extension_management_ui(package_name, title_name, requirements)
-                with gr.Tabs():
-                    main_tab()
+                # with gr.Tabs():
+                main_tab()
         except Exception as e:
             generic_error_tab_advanced(
                 e, name=title_name + " Extension", requirements=requirements