add translation

app.py

@@ -5,20 +5,59 @@ import argparse
 import warnings
 import gradio as gr
 from generate import generate_music, get_args
-from utils import WEIGHTS_DIR, TEMP_DIR
+from utils import WEIGHTS_DIR, TEMP_DIR, LANG
+
+EN2ZH = {
+    "Cite": "引用",
+    "Submit": "提交",
+    "Feedback: the emotion you believe the generated result should belong to": "反馈：你所认为的生成结果该所属的情感",
+    "Status": "状态栏",
+    "Staff": "五线谱",
+    "ABC notation": "ABC 记谱",
+    "Download MXL": "下载 MXL",
+    "Download MusicXML": "下载 MusicXML",
+    "Download PDF score": "下载 PDF 乐谱",
+    "Download MIDI": "下载 MIDI",
+    "Audio": "音频",
+    "Download template": "下载模板",
+    "Save template": "保存模板",
+    "The emotion to which the current template belongs": "当前模板所属情感",
+    "Generate": "生成",
+    "Generate chords coming soon": "生成和声控制暂不可用",
+    "Volume in dB": "dB 音量调节",
+    "±12 octave": "±12 八度上下移",
+    "BPM tempo": "BPM 速度",
+    "Minor": "小调",
+    "Major": "大调",
+    "Mode": "大小调",
+    "Pitch SD": "音高标准差",
+    "Low": "低",
+    "High": "高",
+    "By feature control": "通过特征控制生成",
+    "By template": "通过模板生成",
+    "Arousal: reflects the calmness-intensity of the emotion": "唤醒度 反映情绪的 平静-激烈 程度",
+    "Valence: reflects negative-positive levels of emotion": "愉悦度 反映情绪的 消极-积极 程度",
+    "Video demo": "视频教程",
+    "Dataset": "数据集",
+    "Status": "状态栏",
+}
+
+
+def _L(en_txt: str):
+    return en_txt if LANG else f"{en_txt} ({EN2ZH[en_txt]})"
 
 
 def infer_by_template(dataset: str, v: str, a: str, add_chord: bool):
     status = "Success"
     audio = midi = pdf = xml = mxl = tunes = jpg = None
     emotion = "Q1"
-    if v == "Low" and a == "High":
+    if v == _L("Low") and a == _L("High"):
         emotion = "Q2"
 
-    elif v == "Low" and a == "Low":
+    elif v == _L("Low") and a == _L("Low"):
         emotion = "Q3"
 
-    elif v == "High" and a == "Low":
+    elif v == _L("High") and a == _L("Low"):
         emotion = "Q4"
 
     try:
@@ -49,13 +88,13 @@ def infer_by_features(
     status = "Success"
     audio = midi = pdf = xml = mxl = tunes = jpg = None
     emotion = "Q1"
-    if mode == "Minor" and pitch_std == "High":
+    if mode == _L("Minor") and pitch_std == _L("High"):
         emotion = "Q2"
 
-    elif mode == "Minor" and pitch_std == "Low":
+    elif mode == _L("Minor") and pitch_std == _L("Low"):
         emotion = "Q3"
 
-    elif mode == "Major" and pitch_std == "Low":
+    elif mode == _L("Major") and pitch_std == _L("Low"):
         emotion = "Q4"
 
     try:
@@ -123,8 +162,8 @@ def save_template(label: str, pitch_std: str, mode: str, tempo: int, octave: int
             json_str = json.dumps(
                 {
                     "label": label,
-                    "pitch_std": pitch_std == "High",
-                    "mode": mode == "Major",
+                    "pitch_std": pitch_std == _L("High"),
+                    "mode": mode == _L("Major"),
                     "tempo": tempo,
                     "octave": octave,
                     "volume": rms,
@@ -152,23 +191,24 @@ def save_template(label: str, pitch_std: str, mode: str, tempo: int, octave: int
 if __name__ == "__main__":
     warnings.filterwarnings("ignore")
     with gr.Blocks() as demo:
-        gr.Markdown(
-            "## The current CPU-based version on HuggingFace has slow inference, you can access the GPU-based mirror on [ModelScope](https://www.modelscope.cn/studios/monetjoe/EMelodyGen)"
-        )
+        if LANG:
+            gr.Markdown(
+                "## The current CPU-based version on HuggingFace has slow inference, you can access the GPU-based mirror on [ModelScope](https://www.modelscope.cn/studios/monetjoe/EMelodyGen)"
+            )
         with gr.Row():
             with gr.Column():
                 gr.Video(
-                    "./demo.mp4",
-                    label="Video demo",
+                    "./demo.mp4" if LANG else "./src/tutorial.mp4",
+                    label=_L("Video demo"),
                     show_download_button=False,
                     show_share_button=False,
                 )
                 dataset_option = gr.Dropdown(
                     ["VGMIDI", "EMOPIA", "Rough4Q"],
-                    label="Dataset",
+                    label=_L("Dataset"),
                     value="Rough4Q",
                 )
-                with gr.Tab("By template"):
+                with gr.Tab(_L("By template")):
                     gr.Image(
                         "https://www.modelscope.cn/studio/monetjoe/EMelodyGen/resolve/master/src/4q.jpg",
                         show_label=False,
@@ -177,90 +217,98 @@ if __name__ == "__main__":
                         show_share_button=False,
                     )
                     valence_radio = gr.Radio(
-                        ["Low", "High"],
-                        label="Valence (reflects negative-positive levels of emotion)",
-                        value="High",
+                        [_L("Low"), _L("High")],
+                        label=_L(
+                            "Valence: reflects negative-positive levels of emotion"
+                        ),
+                        value=_L("High"),
                     )
                     arousal_radio = gr.Radio(
-                        ["Low", "High"],
-                        label="Arousal (reflects the calmness-intensity of the emotion)",
-                        value="High",
+                        [_L("Low"), _L("High")],
+                        label=_L(
+                            "Arousal: reflects the calmness-intensity of the emotion"
+                        ),
+                        value=_L("High"),
                     )
                     chord_check = gr.Checkbox(
-                        label="Generate chords (coming soon)",
+                        label=_L("Generate chords coming soon"),
                         value=False,
                     )
-                    gen_btn_1 = gr.Button("Generate")
+                    gen_btn_1 = gr.Button(_L("Generate"))
 
-                with gr.Tab("By feature control"):
+                with gr.Tab(_L("By feature control")):
                     std_option = gr.Radio(
-                        ["Low", "High"], label="Pitch SD", value="High"
+                        [_L("Low"), _L("High")], label=_L("Pitch SD"), value=_L("High")
                     )
                     mode_option = gr.Radio(
-                        ["Minor", "Major"], label="Mode", value="Major"
+                        [_L("Minor"), _L("Major")], label=_L("Mode"), value=_L("Major")
                     )
                     tempo_option = gr.Slider(
                         minimum=40,
                         maximum=228,
                         step=1,
                         value=120,
-                        label="Tempo (BPM)",
+                        label=_L("BPM tempo"),
                     )
                     octave_option = gr.Slider(
                         minimum=-24,
                         maximum=24,
                         step=12,
                         value=0,
-                        label="Octave (±12)",
+                        label=_L("±12 octave"),
                     )
                     volume_option = gr.Slider(
                         minimum=-5,
                         maximum=10,
                         step=5,
                         value=0,
-                        label="Volume (dB)",
+                        label=_L("Volume in dB"),
                     )
                     chord_check_2 = gr.Checkbox(
-                        label="Generate chords (coming soon)",
+                        label=_L("Generate chords coming soon"),
                         value=False,
                     )
-                    gen_btn_2 = gr.Button("Generate")
+                    gen_btn_2 = gr.Button(_L("Generate"))
                     template_radio = gr.Radio(
                         ["Q1", "Q2", "Q3", "Q4"],
-                        label="The emotion to which the current template belongs",
+                        label=_L("The emotion to which the current template belongs"),
                     )
-                    save_btn = gr.Button("Save template")
-                    dld_template = gr.File(label="Download template")
+                    save_btn = gr.Button(_L("Save template"))
+                    dld_template = gr.File(label=_L("Download template"))
 
             with gr.Column():
-                wav_audio = gr.Audio(label="Audio", type="filepath")
-                midi_file = gr.File(label="Download MIDI")
-                pdf_file = gr.File(label="Download PDF score")
-                xml_file = gr.File(label="Download MusicXML")
-                mxl_file = gr.File(label="Download MXL")
-                abc_textbox = gr.Textbox(label="ABC notation", show_copy_button=True)
-                staff_img = gr.Image(label="Staff", type="filepath")
+                wav_audio = gr.Audio(label=_L("Audio"), type="filepath")
+                midi_file = gr.File(label=_L("Download MIDI"))
+                pdf_file = gr.File(label=_L("Download PDF score"))
+                xml_file = gr.File(label=_L("Download MusicXML"))
+                mxl_file = gr.File(label=_L("Download MXL"))
+                abc_textbox = gr.Textbox(
+                    label=_L("ABC notation"), show_copy_button=True
+                )
+                staff_img = gr.Image(label=_L("Staff"), type="filepath")
 
             with gr.Column():
-                status_bar = gr.Textbox(label="Status", show_copy_button=True)
+                status_bar = gr.Textbox(label=_L("Status"), show_copy_button=True)
                 fdb_radio = gr.Radio(
                     ["Q1", "Q2", "Q3", "Q4"],
-                    label="Feedback: the emotion you believe the generated result should belong to",
+                    label=_L(
+                        "Feedback: the emotion you believe the generated result should belong to"
+                    ),
                 )
-                fdb_btn = gr.Button("Submit")
+                fdb_btn = gr.Button(_L("Submit"))
 
                 gr.Markdown(
-                    """## Cite
+                    f"""## {_L("Cite")}
                     ```bibtex
-                    @inproceedings{Zhou2025EMelodyGen,
-                        title     = {EMelodyGen: Emotion-Conditioned Melody Generation in ABC Notation with the Musical Feature Template},
-                        author    = {Monan Zhou and Xiaobing Li and Feng Yu and Wei Li},
-                        month     = {Mar},
-                        year      = {2025},
-                        publisher = {GitHub},
-                        version   = {0.1},
-                        url       = {https://github.com/monetjoe/EMelodyGen}
-                    }
+                    @inproceedings{{Zhou2025EMelodyGen,
+                        title     = {{EMelodyGen: Emotion-Conditioned Melody Generation in ABC Notation with the Musical Feature Template}},
+                        author    = {{Monan Zhou and Xiaobing Li and Feng Yu and Wei Li}},
+                        month     = {{Mar}},
+                        year      = {{2025}},
+                        publisher = {{GitHub}},
+                        version   = {{0.1}},
+                        url       = {{https://github.com/monetjoe/EMelodyGen}}
+                    }}
                     ```"""
                 )
 

utils.py

@@ -9,9 +9,10 @@ import huggingface_hub
 from tqdm import tqdm
 
 TEMP_DIR = "./__pycache__"
+LANG = os.getenv("language")
 WEIGHTS_DIR = (
     huggingface_hub.snapshot_download("monetjoe/EMelodyGen", cache_dir=TEMP_DIR)
-    if os.getenv("language")
+    if LANG
     else modelscope.snapshot_download("monetjoe/EMelodyGen", cache_dir=TEMP_DIR)
 )
 DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")