Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

feat(webui): persist reusable generation settings

HaningZS3f5842bc4835
diff --git a/config.example.toml b/config.example.toml
index 14ced7f9..89783b5f 100644
--- a/config.example.toml
+++ b/config.example.toml
@@ -389,11 +389,34 @@ open_task_folder_on_completion = true
# The WebUI writes the remaining preferences automatically. Uncomment values
# below only when you want to define initial defaults manually.
# language = "zh"
+# video_language = ""
+# paragraph_number = 1
+# video_script_prompt = ""
+# custom_system_prompt = ""
+# loomloom_candidate_count = 3
+# loomloom_script_duration_seconds = 60
+# loomloom_video_scene_count = 1
+# video_concat_mode = "random" # "random" or "sequential"
+# video_transition_mode = "None" # "None", "Shuffle", "FadeIn", etc.
+# video_aspect_pexels = "9:16" # Stored separately for each video source.
+# video_clip_duration = 3
+# video_clip_speed = 1.0
+# video_count = 1
# tts_server = "azure-tts-v1"
# voice_name = ""
+# voice_volume = 1.0
+# voice_rate = 1.0
+# bgm_type = "random"
+# bgm_volume = 0.2
+# custom_bgm_file = ""
+# sonilo_bgm_prompt = ""
+# elevenlabs_music_prompt = ""
+# subtitle_enabled = true
# font_name = "MicrosoftYaHeiBold.ttc"
# font_size = 60
# text_fore_color = "#FFFFFF"
+# stroke_color = "#000000"
+# stroke_width = 1.5
# subtitle_position = "bottom" # "top", "center", "bottom", or "custom"
# custom_position = 70.0 # Percentage from the top when using "custom"
# subtitle_background_enabled = false
diff --git a/test/services/test_webui_bgm.py b/test/services/test_webui_bgm.py
index b507f9f1..7a0f363b 100644
--- a/test/services/test_webui_bgm.py
+++ b/test/services/test_webui_bgm.py
@@ -255,9 +255,13 @@ class TestWebuiBackgroundMusic(unittest.TestCase):
def test_zero_volume_does_not_require_sonilo_key(self):
"""Sonilo 音量为 0 时,WebUI 不应继续显示 API Key 必填警告。"""
test_config = dict(config.app, sonilo_api_key="")
+ # BGM 音量现在是可持久化的用户偏好。显式给定本测试的
+ # 非零初始条件,避免其他 AppTest 会话保存的默认值影响前置断言。
+ test_ui = dict(config.ui, bgm_volume=0.2)
required_warning = self._translation("en", "Sonilo API Key Required")
with (
patch.object(config, "app", test_config),
+ patch.object(config, "ui", test_ui),
patch.object(config, "save_config"),
patch.object(sonilo, "is_enabled", return_value=False),
):
diff --git a/test/services/test_webui_generation_defaults.py b/test/services/test_webui_generation_defaults.py
new file mode 100644
index 00000000..3c9fe5d7
--- /dev/null
+++ b/test/services/test_webui_generation_defaults.py
@@ -0,0 +1,444 @@
+from pathlib import Path
+from unittest.mock import patch
+
+from streamlit.testing.v1 import AppTest
+
+from app.config import config
+from app.services import voice
+
+
+ROOT_DIR = Path(__file__).parent.parent.parent
+WEBUI_MAIN = ROOT_DIR / "webui" / "Main.py"
+
+
+def _widget_by_key(elements, key):
+ return next(
+ item
+ for item in elements
+ if str(getattr(item, "key", "")) == key
+ or str(getattr(item, "key", "")).startswith(f"{key}_")
+ )
+
+
+def _new_app():
+ # A cold Python 3.11 environment can spend over 30 seconds importing the
+ # full Streamlit entrypoint and optional media stack. Keep the assertion
+ # timeout above that one-time startup cost so targeted runs do not flake.
+ app = AppTest.from_file(str(WEBUI_MAIN), default_timeout=60)
+ app.session_state["ui_language"] = "en"
+ app.run()
+ assert [str(item.value) for item in app.exception] == []
+ return app
+
+
+def test_reusable_generation_settings_survive_a_new_webui_session():
+ """Reusable controls should persist while per-video content stays session-only."""
+ test_app_config = dict(
+ config.app,
+ video_source="pexels",
+ match_materials_to_script=False,
+ )
+ test_ui_config = dict(
+ config.ui,
+ language="en",
+ voice_mode="tts",
+ tts_server="azure-tts-v1",
+ voice_name="en-US-JennyNeural-Female",
+ )
+
+ with (
+ patch.object(config, "app", test_app_config),
+ patch.object(config, "ui", test_ui_config),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.object(
+ voice,
+ "get_all_azure_voices",
+ return_value=["en-US-JennyNeural-Female"],
+ ),
+ ):
+ first_session = _new_app()
+ _widget_by_key(first_session.text_area, "video_subject").set_value(
+ "One-off subject"
+ )
+ _widget_by_key(first_session.text_area, "video_script").set_value(
+ "One-off script"
+ )
+ _widget_by_key(first_session.text_area, "video_terms").set_value(
+ "one-off, keywords"
+ )
+ _widget_by_key(first_session.selectbox, "script_language_select").set_value(
+ "en-US"
+ )
+ _widget_by_key(first_session.slider, "paragraph_number_input").set_value(3)
+ _widget_by_key(first_session.text_area, "video_script_prompt").set_value(
+ "Keep the hook concise."
+ )
+ _widget_by_key(first_session.text_area, "custom_system_prompt").set_value(
+ "Write a factual short-video script."
+ )
+ _widget_by_key(first_session.selectbox, "video_concat_mode_select").set_value(
+ "sequential"
+ )
+ _widget_by_key(
+ first_session.selectbox, "video_transition_mode_select"
+ ).set_value("FadeIn")
+ _widget_by_key(first_session.selectbox, "video_aspect_for_pexels").set_value(
+ "16:9"
+ )
+ _widget_by_key(first_session.selectbox, "video_clip_duration_select").set_value(
+ 7
+ )
+ _widget_by_key(first_session.slider, "video_clip_speed_slider").set_value(1.5)
+ _widget_by_key(first_session.selectbox, "video_count_select").set_value(3)
+ _widget_by_key(first_session.selectbox, "voice_volume_select").set_value(1.5)
+ _widget_by_key(first_session.selectbox, "voice_rate_select").set_value(1.2)
+ _widget_by_key(first_session.selectbox, "bgm_type_select").set_value("custom")
+ _widget_by_key(first_session.checkbox, "subtitle_enabled_checkbox").set_value(
+ True
+ )
+ _widget_by_key(first_session.color_picker, "stroke_color_picker").set_value(
+ "#123456"
+ )
+ _widget_by_key(first_session.slider, "stroke_width_slider").set_value(2.5)
+ first_session.run()
+
+ # Aspect is a per-source preference: Coverr's common landscape default
+ # must not replace an explicit portrait choice or the Pexels preference.
+ _widget_by_key(first_session.selectbox, "video_source_select").set_value(
+ "coverr"
+ ).run()
+ _widget_by_key(first_session.selectbox, "video_aspect_for_coverr").set_value(
+ "9:16"
+ ).run()
+ _widget_by_key(first_session.selectbox, "video_source_select").set_value(
+ "pexels"
+ ).run()
+
+ _widget_by_key(first_session.selectbox, "bgm_volume_select").set_value(0.4)
+ _widget_by_key(first_session.text_input, "custom_bgm_file_input").set_value(
+ "example.mp3"
+ )
+ first_session.run()
+ _widget_by_key(first_session.selectbox, "bgm_type_select").set_value(
+ "sonilo"
+ ).run()
+ _widget_by_key(
+ first_session.text_input, "sonilo_bgm_prompt_input"
+ ).set_value("Bright acoustic underscore").run()
+ _widget_by_key(first_session.selectbox, "bgm_type_select").set_value(
+ "elevenlabs"
+ ).run()
+ _widget_by_key(
+ first_session.text_input, "elevenlabs_music_prompt_input"
+ ).set_value("Calm cinematic underscore").run()
+ _widget_by_key(first_session.selectbox, "bgm_type_select").set_value(
+ "custom"
+ ).run()
+ assert [str(item.value) for item in first_session.exception] == []
+
+ expected_defaults = {
+ "video_language": "en-US",
+ "paragraph_number": 3,
+ "video_script_prompt": "Keep the hook concise.",
+ "custom_system_prompt": "Write a factual short-video script.",
+ "video_concat_mode": "sequential",
+ "video_transition_mode": "FadeIn",
+ "video_aspect_pexels": "16:9",
+ "video_aspect_coverr": "9:16",
+ "video_clip_duration": 7,
+ "video_clip_speed": 1.5,
+ "video_count": 3,
+ "voice_volume": 1.5,
+ "voice_rate": 1.2,
+ "bgm_type": "custom",
+ "bgm_volume": 0.4,
+ "custom_bgm_file": "example.mp3",
+ "sonilo_bgm_prompt": "Bright acoustic underscore",
+ "elevenlabs_music_prompt": "Calm cinematic underscore",
+ "subtitle_enabled": True,
+ "stroke_color": "#123456",
+ "stroke_width": 2.5,
+ }
+ assert {key: test_ui_config.get(key) for key in expected_defaults} == (
+ expected_defaults
+ )
+ assert "video_subject" not in test_ui_config
+ assert "video_script" not in test_ui_config
+ assert "video_terms" not in test_ui_config
+ assert "video_subject" not in test_app_config
+ assert "video_script" not in test_app_config
+ assert "video_terms" not in test_app_config
+
+ second_session = _new_app()
+ assert _widget_by_key(
+ second_session.selectbox, "script_language_select"
+ ).value == "en-US"
+ assert _widget_by_key(
+ second_session.slider, "paragraph_number_input"
+ ).value == 3
+ assert _widget_by_key(
+ second_session.text_area, "video_script_prompt"
+ ).value == "Keep the hook concise."
+ assert _widget_by_key(
+ second_session.text_area, "custom_system_prompt"
+ ).value == "Write a factual short-video script."
+ assert _widget_by_key(
+ second_session.selectbox, "video_concat_mode_select"
+ ).value == "sequential"
+ assert _widget_by_key(
+ second_session.selectbox, "video_transition_mode_select"
+ ).value == "FadeIn"
+ assert _widget_by_key(
+ second_session.selectbox, "video_aspect_for_pexels"
+ ).value == "16:9"
+ _widget_by_key(second_session.selectbox, "video_source_select").set_value(
+ "coverr"
+ ).run()
+ assert _widget_by_key(
+ second_session.selectbox, "video_aspect_for_coverr"
+ ).value == "9:16"
+ _widget_by_key(second_session.selectbox, "video_source_select").set_value(
+ "pexels"
+ ).run()
+ assert _widget_by_key(
+ second_session.selectbox, "video_clip_duration_select"
+ ).value == 7
+ assert _widget_by_key(
+ second_session.slider, "video_clip_speed_slider"
+ ).value == 1.5
+ assert _widget_by_key(second_session.selectbox, "video_count_select").value == 3
+ assert _widget_by_key(
+ second_session.selectbox, "voice_volume_select"
+ ).value == 1.5
+ assert _widget_by_key(
+ second_session.selectbox, "voice_rate_select"
+ ).value == 1.2
+ assert _widget_by_key(second_session.selectbox, "bgm_type_select").value == (
+ "custom"
+ )
+ assert _widget_by_key(second_session.selectbox, "bgm_volume_select").value == (
+ 0.4
+ )
+ assert _widget_by_key(
+ second_session.text_input, "custom_bgm_file_input"
+ ).value == "example.mp3"
+ _widget_by_key(second_session.selectbox, "bgm_type_select").set_value(
+ "sonilo"
+ ).run()
+ assert _widget_by_key(
+ second_session.text_input, "sonilo_bgm_prompt_input"
+ ).value == "Bright acoustic underscore"
+ _widget_by_key(second_session.selectbox, "bgm_type_select").set_value(
+ "elevenlabs"
+ ).run()
+ assert _widget_by_key(
+ second_session.text_input, "elevenlabs_music_prompt_input"
+ ).value == "Calm cinematic underscore"
+ assert _widget_by_key(
+ second_session.checkbox, "subtitle_enabled_checkbox"
+ ).value is True
+ assert _widget_by_key(
+ second_session.color_picker, "stroke_color_picker"
+ ).value == "#123456"
+ assert _widget_by_key(second_session.slider, "stroke_width_slider").value == 2.5
+
+ # Per-video content must not leak into a new session.
+ assert second_session.session_state["video_subject"] == ""
+ assert second_session.session_state["video_script"] == ""
+ assert second_session.session_state["video_terms"] == ""
+
+ _widget_by_key(
+ second_session.button, "restore_default_system_prompt"
+ ).click().run()
+ assert test_ui_config["custom_system_prompt"] == ""
+
+ _widget_by_key(
+ second_session.button, "restore_default_subtitle_settings"
+ ).click().run()
+ assert {
+ key: test_ui_config[key]
+ for key in ("subtitle_enabled", "stroke_color", "stroke_width")
+ } == {
+ "subtitle_enabled": True,
+ "stroke_color": "#000000",
+ "stroke_width": 1.5,
+ }
+
+
+def test_invalid_saved_generation_settings_fall_back_without_breaking_webui():
+ """Stale or manually edited TOML values must not poison Streamlit widgets."""
+ test_app_config = dict(
+ config.app,
+ video_source="pexels",
+ match_materials_to_script=False,
+ )
+ test_ui_config = dict(
+ config.ui,
+ language="en",
+ voice_mode="tts",
+ tts_server="azure-tts-v1",
+ voice_name="en-US-JennyNeural-Female",
+ video_language="not-a-language",
+ paragraph_number=999,
+ video_concat_mode="not-a-mode",
+ video_transition_mode="not-a-transition",
+ video_aspect_pexels="4:3",
+ video_clip_duration=999,
+ video_clip_speed="nan",
+ video_count=True,
+ voice_volume=999,
+ voice_rate=-1,
+ bgm_type="not-a-source",
+ bgm_volume=999,
+ subtitle_enabled="false",
+ stroke_color="not-a-color",
+ stroke_width="inf",
+ loomloom_candidate_count=0,
+ loomloom_script_duration_seconds=9999,
+ loomloom_video_scene_count="nan",
+ )
+
+ with (
+ patch.object(config, "app", test_app_config),
+ patch.object(config, "ui", test_ui_config),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.object(
+ voice,
+ "get_all_azure_voices",
+ return_value=["en-US-JennyNeural-Female"],
+ ),
+ ):
+ app = _new_app()
+
+ assert _widget_by_key(app.selectbox, "script_language_select").value == ""
+ assert _widget_by_key(app.slider, "paragraph_number_input").value == 10
+ assert _widget_by_key(app.selectbox, "video_concat_mode_select").value == "random"
+ assert _widget_by_key(app.selectbox, "video_transition_mode_select").value == (
+ "None"
+ )
+ assert _widget_by_key(app.selectbox, "video_aspect_for_pexels").value == "9:16"
+ assert _widget_by_key(app.selectbox, "video_clip_duration_select").value == 3
+ assert _widget_by_key(app.slider, "video_clip_speed_slider").value == 1.0
+ assert _widget_by_key(app.selectbox, "video_count_select").value == 1
+ assert isinstance(test_ui_config["video_count"], int)
+ assert not isinstance(test_ui_config["video_count"], bool)
+ assert _widget_by_key(app.selectbox, "voice_volume_select").value == 1.0
+ assert _widget_by_key(app.selectbox, "voice_rate_select").value == 1.0
+ assert _widget_by_key(app.selectbox, "bgm_type_select").value == "random"
+ assert _widget_by_key(app.selectbox, "bgm_volume_select").value == 0.2
+ assert _widget_by_key(app.checkbox, "subtitle_enabled_checkbox").value is False
+ assert _widget_by_key(app.color_picker, "stroke_color_picker").value == "#000000"
+ assert _widget_by_key(app.slider, "stroke_width_slider").value == 1.5
+ assert app.session_state["loomloom_candidate_count"] == 1
+ assert app.session_state["loomloom_script_duration_seconds"] == 600
+ assert app.session_state["loomloom_video_scene_count"] == 1
+
+
+def test_loomloom_tuning_survives_restart_without_persisting_payment_state():
+ """Paid-provider tuning is reusable; quotes and confirmations are not."""
+ test_app_config = dict(
+ config.app,
+ video_source="loomloom",
+ script_generation_backend="loomloom",
+ loomloom_api_token="",
+ match_materials_to_script=False,
+ )
+ test_ui_config = dict(
+ config.ui,
+ language="en",
+ voice_mode="tts",
+ tts_server="azure-tts-v1",
+ voice_name="en-US-JennyNeural-Female",
+ )
+
+ with (
+ patch.object(config, "app", test_app_config),
+ patch.object(config, "ui", test_ui_config),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.object(
+ voice,
+ "get_all_azure_voices",
+ return_value=["en-US-JennyNeural-Female"],
+ ),
+ ):
+ first_session = _new_app()
+ _widget_by_key(
+ first_session.number_input, "loomloom_candidate_count"
+ ).set_value(4)
+ _widget_by_key(
+ first_session.number_input, "loomloom_script_duration_seconds"
+ ).set_value(120)
+ _widget_by_key(
+ first_session.number_input, "loomloom_video_scene_count"
+ ).set_value(3)
+ first_session.run()
+
+ assert {
+ key: test_ui_config.get(key)
+ for key in (
+ "loomloom_candidate_count",
+ "loomloom_script_duration_seconds",
+ "loomloom_video_scene_count",
+ )
+ } == {
+ "loomloom_candidate_count": 4,
+ "loomloom_script_duration_seconds": 120,
+ "loomloom_video_scene_count": 3,
+ }
+ assert "loomloom_confirm_charge" not in test_ui_config
+ assert "loomloom_video_confirm_charge" not in test_ui_config
+
+ second_session = _new_app()
+ assert _widget_by_key(
+ second_session.number_input, "loomloom_candidate_count"
+ ).value == 4
+ assert _widget_by_key(
+ second_session.number_input, "loomloom_script_duration_seconds"
+ ).value == 120
+ assert _widget_by_key(
+ second_session.number_input, "loomloom_video_scene_count"
+ ).value == 3
+ assert second_session.session_state["loomloom_video_confirm_charge"] is False
+
+
+def test_script_order_constraint_does_not_replace_saved_concat_preference():
+ """A derived sequential mode must not become the user's reusable default."""
+ test_app_config = dict(
+ config.app,
+ video_source="pexels",
+ match_materials_to_script=True,
+ )
+ test_ui_config = dict(
+ config.ui,
+ language="en",
+ video_concat_mode="random",
+ voice_mode="tts",
+ tts_server="azure-tts-v1",
+ voice_name="en-US-JennyNeural-Female",
+ )
+
+ with (
+ patch.object(config, "app", test_app_config),
+ patch.object(config, "ui", test_ui_config),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.object(
+ voice,
+ "get_all_azure_voices",
+ return_value=["en-US-JennyNeural-Female"],
+ ),
+ ):
+ constrained_session = _new_app()
+ constrained_concat = _widget_by_key(
+ constrained_session.selectbox, "video_concat_mode_select"
+ )
+ assert constrained_concat.value == "sequential"
+ assert constrained_concat.disabled is True
+ assert test_ui_config["video_concat_mode"] == "random"
+
+ # A later session without the constraint should recover the user's actual
+ # preference, not the derived value shown while script-order matching ran.
+ test_app_config["match_materials_to_script"] = False
+ unconstrained_session = _new_app()
+ assert _widget_by_key(
+ unconstrained_session.selectbox, "video_concat_mode_select"
+ ).value == "random"
diff --git a/webui/Main.py b/webui/Main.py
index 50e154bf..0ea8e6a8 100644
--- a/webui/Main.py
+++ b/webui/Main.py
@@ -178,6 +178,79 @@ def _save_runtime_config():
return saved
+def _saved_ui_choice(key, options, default):
+ """读取一个持久化选择,并把旧配置或手工编辑的非法值降级为默认值。"""
+ options = list(options)
+ saved = config.ui.get(key, default)
+ numeric_default = isinstance(default, (int, float)) and not isinstance(
+ default, bool
+ )
+ # bool 是 int 的子类,``True == 1``。手工把数值选项写成 TOML
+ # 布尔值时必须拒绝,不能让它伪装成第一个数值 option。
+ if numeric_default and isinstance(saved, bool):
+ return default
+ for option in options:
+ if saved == option:
+ # 返回 options 中的真实值,顺便把 TOML 1.0 等价归一化为
+ # 整数选项 1,避免下游参数类型随配置写法漂移。
+ return option
+
+ # TOML 中的数值通常保留原类型;仍兼容用户手工写成字符串的情况。
+ if numeric_default and isinstance(saved, str):
+ try:
+ converted = type(default)(saved)
+ except (TypeError, ValueError):
+ converted = None
+ for option in options:
+ if converted == option:
+ return option
+ return default
+
+
+def _saved_ui_number(key, default, minimum, maximum, number_type=float):
+ """读取并限幅持久化数值,避免非法配置破坏 Streamlit slider。"""
+ try:
+ saved = config.ui.get(key, default)
+ if isinstance(saved, bool):
+ raise ValueError("boolean is not a numeric setting")
+ value = number_type(saved)
+ if isinstance(value, float) and not math.isfinite(value):
+ raise ValueError("non-finite value")
+ except (TypeError, ValueError, OverflowError):
+ value = default
+ return min(maximum, max(minimum, value))
+
+
+def _saved_ui_bool(key, default):
+ """兼容 TOML 布尔值和常见手工字符串,拒绝含义不明的旧值。"""
+ value = config.ui.get(key, default)
+ if isinstance(value, bool):
+ return value
+ if isinstance(value, str):
+ normalized = value.strip().lower()
+ if normalized in {"1", "true", "yes", "on"}:
+ return True
+ if normalized in {"0", "false", "no", "off"}:
+ return False
+ return default
+
+
+def _saved_ui_color(key, default):
+ """只把标准六位十六进制颜色传给 Streamlit color picker。"""
+ value = str(config.ui.get(key, default) or "").strip()
+ if re.fullmatch(r"#[0-9a-fA-F]{6}", value):
+ return value
+ return default
+
+
+def _saved_ui_text(key, default="", max_length=None):
+ """读取持久化文本并遵守对应 WebUI 控件的长度上限。"""
+ value = str(config.ui.get(key, default) or default)
+ if max_length is not None:
+ value = value[:max_length]
+ return value
+
+
def _run_llm_read_operation(operation_name, operation):
"""
使用稳定的当前 LLM 配置执行只读请求,并避免等待视频生成任务。
@@ -325,11 +398,49 @@ def _initialize_session_state():
"video_subject": "",
"video_script": "",
"video_terms": "",
- "video_script_prompt": "",
- "custom_system_prompt": llm.DEFAULT_SCRIPT_SYSTEM_PROMPT,
+ "paragraph_number_input": _saved_ui_number(
+ "paragraph_number",
+ 1,
+ llm.MIN_SCRIPT_PARAGRAPH_NUMBER,
+ llm.MAX_SCRIPT_PARAGRAPH_NUMBER,
+ int,
+ ),
+ "video_script_prompt": _saved_ui_text(
+ "video_script_prompt",
+ max_length=llm.MAX_SCRIPT_PROMPT_LENGTH,
+ ),
+ "custom_system_prompt": _saved_ui_text(
+ "custom_system_prompt",
+ llm.DEFAULT_SCRIPT_SYSTEM_PROMPT,
+ llm.MAX_SCRIPT_SYSTEM_PROMPT_LENGTH,
+ ),
"match_materials_to_script": bool(
config.app.get("match_materials_to_script", False)
),
+ "custom_bgm_file_input": _saved_ui_text("custom_bgm_file"),
+ "sonilo_bgm_prompt_input": _saved_ui_text(
+ "sonilo_bgm_prompt",
+ max_length=sonilo_service.MAX_PROMPT_LENGTH,
+ ),
+ "elevenlabs_music_prompt_input": _saved_ui_text(
+ "elevenlabs_music_prompt",
+ max_length=elevenlabs_music_service.MAX_PROMPT_LENGTH,
+ ),
+ "subtitle_enabled_checkbox": _saved_ui_bool("subtitle_enabled", True),
+ "stroke_color_picker": _saved_ui_color("stroke_color", "#000000"),
+ "stroke_width_slider": _saved_ui_number(
+ "stroke_width", 1.5, 0.0, 10.0
+ ),
+ "loomloom_candidate_count": _saved_ui_number(
+ "loomloom_candidate_count",
+ 3,
+ 1,
+ loomloom.MAX_SCRIPT_CANDIDATES,
+ int,
+ ),
+ "loomloom_script_duration_seconds": _saved_ui_number(
+ "loomloom_script_duration_seconds", 60, 10, 600, int
+ ),
"ui_language": initial_ui_language,
# 已落盘的本地素材允许用户只修改文案后继续复用。
"local_video_materials": [],
@@ -359,7 +470,13 @@ def _initialize_session_state():
"loomloom_video_client_request_id": "",
"loomloom_video_confirm_charge": False,
# AI 视频按素材段计费,默认只生成一段,用户确认效果后再主动增加数量。
- "loomloom_video_scene_count": 1,
+ "loomloom_video_scene_count": _saved_ui_number(
+ "loomloom_video_scene_count",
+ 1,
+ 1,
+ loomloom.MAX_VIDEO_SCENES,
+ int,
+ ),
}
for key, value in defaults.items():
st.session_state.setdefault(key, value)
@@ -1744,11 +1861,14 @@ def reset_subtitle_settings():
# 同步会持久化的 UI 选项,确保恢复后刷新页面仍保持默认设置。
for key in (
+ "subtitle_enabled",
"font_name",
"subtitle_position",
"custom_position",
"text_fore_color",
"font_size",
+ "stroke_color",
+ "stroke_width",
"subtitle_background_enabled",
"subtitle_background_color",
"rounded_subtitle_background",
@@ -2372,13 +2492,14 @@ def _render_loomloom_video_settings(params):
token = _effective_loomloom_api_token()
- st.number_input(
+ scene_count = st.number_input(
tr("AI Video Scene Count"),
min_value=1,
max_value=loomloom.MAX_VIDEO_SCENES,
step=1,
key="loomloom_video_scene_count",
)
+ _set_runtime_config("ui", "loomloom_video_scene_count", int(scene_count))
batch, input_signature = _current_loomloom_video_quote_context(params)
if not token:
st.warning(tr("Shengsuan Cloud API Key Required"))
@@ -2641,7 +2762,6 @@ def _render_loomloom_script_generation(params):
tr("Script Candidate Count"),
min_value=1,
max_value=loomloom.MAX_SCRIPT_CANDIDATES,
- value=3,
step=1,
key="loomloom_candidate_count",
)
@@ -2649,10 +2769,13 @@ def _render_loomloom_script_generation(params):
tr("Target Script Duration Seconds"),
min_value=10,
max_value=600,
- value=60,
step=10,
key="loomloom_script_duration_seconds",
)
+ _set_runtime_config("ui", "loomloom_candidate_count", int(candidate_count))
+ _set_runtime_config(
+ "ui", "loomloom_script_duration_seconds", int(duration_seconds)
+ )
input_signature = _loomloom_script_signature(
subject=params.video_subject,
language=params.video_language,
@@ -2831,13 +2954,18 @@ def _render_script_settings(panel, params):
selected_language_code = stable_selectbox(
tr("Script Language"),
options=[value for _, value in video_languages],
- default_value="",
+ default_value=_saved_ui_choice(
+ "video_language",
+ [value for _, value in video_languages],
+ "",
+ ),
key="script_language_select",
format_func=lambda value: dict(
(v, label) for label, v in video_languages
)[value],
)
params.video_language = selected_language_code
+ _set_runtime_config("ui", "video_language", params.video_language)
# 使用带 key 的局部容器限定折叠入口样式,保持 expander 的原生交互,
# 同时避免样式误伤页面顶部的“基础设置”等其他折叠区域。
@@ -2860,13 +2988,15 @@ def _render_script_settings(panel, params):
"app", "script_generation_backend", script_generation_backend
)
- st.session_state.setdefault("paragraph_number_input", 1)
params.paragraph_number = st.slider(
tr("Script Paragraph Number"),
min_value=llm.MIN_SCRIPT_PARAGRAPH_NUMBER,
max_value=llm.MAX_SCRIPT_PARAGRAPH_NUMBER,
key="paragraph_number_input",
)
+ _set_runtime_config(
+ "ui", "paragraph_number", params.paragraph_number
+ )
params.video_script_prompt = st.text_area(
tr("Custom Script Requirements"),
height=100,
@@ -2874,6 +3004,9 @@ def _render_script_settings(panel, params):
placeholder=tr("Custom Script Requirements Placeholder"),
key="video_script_prompt",
).strip()
+ _set_runtime_config(
+ "ui", "video_script_prompt", params.video_script_prompt
+ )
system_prompt = st.text_area(
tr("Custom System Prompt"),
@@ -2888,6 +3021,9 @@ def _render_script_settings(panel, params):
if system_prompt == llm.DEFAULT_SCRIPT_SYSTEM_PROMPT.strip()
else system_prompt
)
+ _set_runtime_config(
+ "ui", "custom_system_prompt", params.custom_system_prompt
+ )
restore_prompt_col, preview_prompt_col = st.columns(2)
if restore_prompt_col.button(
@@ -3016,7 +3152,11 @@ def _render_video_settings(panel, params):
selected_concat_mode = stable_selectbox(
tr("Video Concat Mode"),
options=[value for _, value in video_concat_modes],
- default_value=VideoConcatMode.random.value,
+ default_value=_saved_ui_choice(
+ "video_concat_mode",
+ [value for _, value in video_concat_modes],
+ VideoConcatMode.random.value,
+ ),
key="video_concat_mode_select",
format_func=lambda value: dict(
(v, label) for label, v in video_concat_modes
@@ -3036,6 +3176,12 @@ def _render_video_settings(panel, params):
"match_materials_to_script",
params.match_materials_to_script,
)
+ # 顺序匹配开启时,sequential 是派生出的强制值,不应覆盖用户在关闭
+ # 该功能时选择的拼接偏好;关闭后仍能恢复此前的 random/sequential。
+ if not params.match_materials_to_script:
+ _set_runtime_config(
+ "ui", "video_concat_mode", params.video_concat_mode.value
+ )
# 视频转场模式
video_transition_modes = [
@@ -3051,13 +3197,22 @@ def _render_video_settings(panel, params):
selected_transition_mode = stable_selectbox(
tr("Video Transition Mode"),
options=[value for _, value in video_transition_modes],
- default_value=VideoTransitionMode.none.value,
+ default_value=_saved_ui_choice(
+ "video_transition_mode",
+ [value for _, value in video_transition_modes],
+ VideoTransitionMode.none.value,
+ ),
key="video_transition_mode_select",
format_func=lambda value: dict(
(v, label) for label, v in video_transition_modes
)[value],
)
params.video_transition_mode = VideoTransitionMode(selected_transition_mode)
+ _set_runtime_config(
+ "ui",
+ "video_transition_mode",
+ params.video_transition_mode.value,
+ )
video_aspect_ratios = [
(tr("Portrait"), VideoAspect.portrait.value),
@@ -3070,30 +3225,48 @@ def _render_video_settings(panel, params):
# - 用户在某 source 下手动改过 aspect,session_state 会记住,
# 下次回到同一 source 时尊重用户选择,不会再被强制覆盖。
default_aspect_index = 1 if params.video_source == "coverr" else 0
+ video_aspect_values = [value for _, value in video_aspect_ratios]
+ video_aspect_config_key = f"video_aspect_{params.video_source}"
selected_aspect_ratio = stable_selectbox(
tr("Video Ratio"),
- options=[value for _, value in video_aspect_ratios],
- default_value=video_aspect_ratios[default_aspect_index][1],
+ options=video_aspect_values,
+ default_value=_saved_ui_choice(
+ video_aspect_config_key,
+ video_aspect_values,
+ video_aspect_ratios[default_aspect_index][1],
+ ),
key=f"video_aspect_for_{params.video_source}",
format_func=lambda value: dict(
(v, label) for label, v in video_aspect_ratios
)[value],
)
params.video_aspect = VideoAspect(selected_aspect_ratio)
+ _set_runtime_config(
+ "ui", video_aspect_config_key, params.video_aspect.value
+ )
+ video_clip_durations = [2, 3, 4, 5, 6, 7, 8, 9, 10]
params.video_clip_duration = stable_selectbox(
tr("Clip Duration"),
- options=[2, 3, 4, 5, 6, 7, 8, 9, 10],
- default_value=3,
+ options=video_clip_durations,
+ default_value=_saved_ui_choice(
+ "video_clip_duration", video_clip_durations, 3
+ ),
key="video_clip_duration_select",
help=tr("Clip Duration Help"),
)
+ _set_runtime_config(
+ "ui", "video_clip_duration", params.video_clip_duration
+ )
clip_speed_key = localized_widget_key("video_clip_speed_slider")
# session_state 可能来自旧任务、API 参数或旧版页面状态。控件创建前
# 统一归一化,既保留合法选择,也确保 slider 始终收到 0.5~2.0
# 范围内的有限浮点数。
st.session_state[clip_speed_key] = utils.normalize_clip_speed(
- st.session_state.get(clip_speed_key, 1.0)
+ st.session_state.get(
+ clip_speed_key,
+ _saved_ui_number("video_clip_speed", 1.0, 0.5, 2.0),
+ )
)
params.video_clip_speed = st.slider(
tr("Clip Speed"),
@@ -3104,12 +3277,17 @@ def _render_video_settings(panel, params):
key=clip_speed_key,
help=tr("Clip Speed Help"),
)
+ _set_runtime_config("ui", "video_clip_speed", params.video_clip_speed)
+ video_count_options = [1, 2, 3, 4, 5]
params.video_count = stable_selectbox(
tr("Number of Videos Generated Simultaneously"),
- options=[1, 2, 3, 4, 5],
- default_value=1,
+ options=video_count_options,
+ default_value=_saved_ui_choice(
+ "video_count", video_count_options, 1
+ ),
key="video_count_select",
)
+ _set_runtime_config("ui", "video_count", params.video_count)
video_codec_options = [
(tr("Default Video Encoder"), DEFAULT_VIDEO_CODEC_OPTION),
@@ -3740,6 +3918,7 @@ def _render_elevenlabs_api_key_input(label_key):
def _render_background_music_settings(params, elevenlabs_api_key_rendered=False):
"""渲染背景音乐来源与音量设置,并返回本次待保存的上传文件。"""
uploaded_bgm_file = None
+ previous_bgm_type = st.session_state.get("last_rendered_bgm_type")
st.divider()
bgm_options = [
(tr("No Background Music"), ""),
@@ -3751,11 +3930,16 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
selected_bgm_type = stable_selectbox(
tr("Background Music Source"),
options=[value for _, value in bgm_options],
- default_value="random",
+ default_value=_saved_ui_choice(
+ "bgm_type",
+ [value for _, value in bgm_options],
+ "random",
+ ),
key="bgm_type_select",
format_func=lambda value: dict((v, label) for label, v in bgm_options)[value],
)
params.bgm_type = selected_bgm_type
+ _set_runtime_config("ui", "bgm_type", params.bgm_type)
if params.bgm_type == "sonilo":
configured_key = str(config.app.get("sonilo_api_key", "") or "").strip()
effective_key = configured_key or os.getenv("SONILO_API_KEY", "").strip()
@@ -3778,14 +3962,16 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
else:
_render_elevenlabs_api_key_input("ElevenLabs Music API Key")
+ bgm_volume_options = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]
params.bgm_volume = stable_selectbox(
tr("Background Music Volume"),
- options=[0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0],
- default_value=0.2,
+ options=bgm_volume_options,
+ default_value=_saved_ui_choice("bgm_volume", bgm_volume_options, 0.2),
key="bgm_volume_select",
format_func=lambda value: f"{int(value * 100)}%",
disabled=not params.bgm_type,
)
+ _set_runtime_config("ui", "bgm_volume", params.bgm_volume)
bgm_enabled = bgm_service.should_use_bgm(params.bgm_type, params.bgm_volume)
if params.bgm_type == "custom":
@@ -3878,11 +4064,21 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
st.info(f"{tr('Background Music Ready')}: {safe_name}")
params.bgm_file = safe_name
+ # Streamlit 会在条件控件暂时不渲染时清理其 widget state。
+ # 从其它 BGM 来源切回时用已持久化值恢复;同一来源下
+ # 用户主动清空时 previous_bgm_type 不变,因此不会被旧值反弹。
+ if previous_bgm_type != "custom":
+ st.session_state["custom_bgm_file_input"] = _saved_ui_text(
+ "custom_bgm_file"
+ )
custom_bgm_file = st.text_input(
tr("Custom Background Music File"),
key="custom_bgm_file_input",
disabled=uploaded_bgm_file is not None,
)
+ _set_runtime_config(
+ "ui", "custom_bgm_file", custom_bgm_file.strip()
+ )
if uploaded_bgm_file is None and custom_bgm_file and bgm_enabled:
# 文件名由服务层映射到 storage/bgm 或 resource/songs 后校验,
# UI 不接受两个白名单目录之外的任意路径。
@@ -3893,12 +4089,20 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
params.bgm_file = ""
if params.bgm_type == "sonilo":
+ if previous_bgm_type != "sonilo":
+ st.session_state["sonilo_bgm_prompt_input"] = _saved_ui_text(
+ "sonilo_bgm_prompt",
+ max_length=sonilo_service.MAX_PROMPT_LENGTH,
+ )
params.video_music_prompt = st.text_input(
tr("Sonilo Music Prompt"),
key="sonilo_bgm_prompt_input",
max_chars=sonilo_service.MAX_PROMPT_LENGTH,
help=tr("Sonilo Music Prompt Help"),
).strip()
+ _set_runtime_config(
+ "ui", "sonilo_bgm_prompt", params.video_music_prompt
+ )
if params.video_count > 1:
st.warning(tr("Sonilo Multiple Videos Warning"))
if st.button(
@@ -3914,12 +4118,20 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
else:
st.success(tr("Sonilo Connection Test Succeeded"))
elif params.bgm_type == "elevenlabs":
+ if previous_bgm_type != "elevenlabs":
+ st.session_state["elevenlabs_music_prompt_input"] = _saved_ui_text(
+ "elevenlabs_music_prompt",
+ max_length=elevenlabs_music_service.MAX_PROMPT_LENGTH,
+ )
params.video_music_prompt = st.text_input(
tr("ElevenLabs Music Prompt"),
key="elevenlabs_music_prompt_input",
max_chars=elevenlabs_music_service.MAX_PROMPT_LENGTH,
help=tr("ElevenLabs Music Prompt Help"),
).strip()
+ _set_runtime_config(
+ "ui", "elevenlabs_music_prompt", params.video_music_prompt
+ )
if params.video_count > 1:
st.warning(tr("ElevenLabs Multiple Videos Warning"))
if st.button(
@@ -3946,6 +4158,7 @@ def _render_background_music_settings(params, elevenlabs_api_key_rendered=False)
and not elevenlabs_music_service.is_enabled()
):
st.warning(tr("ElevenLabs API Key Required"))
+ st.session_state["last_rendered_bgm_type"] = params.bgm_type
return uploaded_bgm_file
@@ -4319,14 +4532,18 @@ def _render_audio_settings(panel, params):
params.voice_volume = 1.0
params.voice_rate = 1.0
uploaded_audio_file = None
+ voice_volume_options = [0.6, 0.8, 1.0, 1.2, 1.5, 2.0, 3.0, 4.0, 5.0]
+ voice_rate_options = [0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.5, 1.8, 2.0]
if tts_mode_enabled:
voice_control_cols = st.columns(2)
with voice_control_cols[0]:
params.voice_volume = stable_selectbox(
tr("Voiceover Volume"),
- options=[0.6, 0.8, 1.0, 1.2, 1.5, 2.0, 3.0, 4.0, 5.0],
- default_value=1.0,
+ options=voice_volume_options,
+ default_value=_saved_ui_choice(
+ "voice_volume", voice_volume_options, 1.0
+ ),
key="voice_volume_select",
format_func=lambda value: f"{int(value * 100)}%",
help=tr("Voiceover Volume Help"),
@@ -4335,12 +4552,16 @@ def _render_audio_settings(panel, params):
with voice_control_cols[1]:
params.voice_rate = stable_selectbox(
tr("Voiceover Speed"),
- options=[0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.5, 1.8, 2.0],
- default_value=1.0,
+ options=voice_rate_options,
+ default_value=_saved_ui_choice(
+ "voice_rate", voice_rate_options, 1.0
+ ),
key="voice_rate_select",
format_func=lambda value: f"{value:.1f}×",
help=tr("Voiceover Speed Help"),
)
+ _set_runtime_config("ui", "voice_volume", params.voice_volume)
+ _set_runtime_config("ui", "voice_rate", params.voice_rate)
# 试听必须位于音量和语速控件之后,确保调用使用当前控件值。
_render_voice_preview(
@@ -4363,12 +4584,15 @@ def _render_audio_settings(panel, params):
)
params.voice_volume = stable_selectbox(
tr("Voiceover Volume"),
- options=[0.6, 0.8, 1.0, 1.2, 1.5, 2.0, 3.0, 4.0, 5.0],
- default_value=1.0,
+ options=voice_volume_options,
+ default_value=_saved_ui_choice(
+ "voice_volume", voice_volume_options, 1.0
+ ),
key="voice_volume_select",
format_func=lambda value: f"{int(value * 100)}%",
help=tr("Voiceover Volume Help"),
)
+ _set_runtime_config("ui", "voice_volume", params.voice_volume)
if uploaded_audio_file:
st.audio(uploaded_audio_file, format="audio/mp3")
st.info(
@@ -4390,12 +4614,16 @@ def _render_subtitle_settings(panel, params):
st.write(tr("Subtitle Settings"))
st.session_state.setdefault(
"subtitle_enabled_checkbox",
- DEFAULT_SUBTITLE_SETTINGS["subtitle_enabled"],
+ _saved_ui_bool(
+ "subtitle_enabled",
+ DEFAULT_SUBTITLE_SETTINGS["subtitle_enabled"],
+ ),
)
params.subtitle_enabled = st.checkbox(
tr("Enable Subtitles"),
key="subtitle_enabled_checkbox",
)
+ _set_runtime_config("ui", "subtitle_enabled", params.subtitle_enabled)
subtitle_settings_disabled = not params.subtitle_enabled
font_names = get_all_fonts()
saved_font_name = config.ui.get(
@@ -4495,16 +4723,26 @@ def _render_subtitle_settings(panel, params):
stroke_cols = st.columns([0.42, 0.58])
with stroke_cols[0]:
st.session_state.setdefault(
- "stroke_color_picker", DEFAULT_SUBTITLE_SETTINGS["stroke_color"]
+ "stroke_color_picker",
+ _saved_ui_color(
+ "stroke_color", DEFAULT_SUBTITLE_SETTINGS["stroke_color"]
+ ),
)
params.stroke_color = st.color_picker(
tr("Stroke Color"),
key="stroke_color_picker",
disabled=subtitle_settings_disabled,
)
+ _set_runtime_config("ui", "stroke_color", params.stroke_color)
with stroke_cols[1]:
st.session_state.setdefault(
- "stroke_width_slider", DEFAULT_SUBTITLE_SETTINGS["stroke_width"]
+ "stroke_width_slider",
+ _saved_ui_number(
+ "stroke_width",
+ DEFAULT_SUBTITLE_SETTINGS["stroke_width"],
+ 0.0,
+ 10.0,
+ ),
)
params.stroke_width = st.slider(
tr("Stroke Width"),
@@ -4513,6 +4751,7 @@ def _render_subtitle_settings(panel, params):
key="stroke_width_slider",
disabled=subtitle_settings_disabled,
)
+ _set_runtime_config("ui", "stroke_width", params.stroke_width)
# 背景开关的本地化名称普遍比颜色标签更长,因此让开关占据略多空间。
subtitle_bg_cols = st.columns([0.55, 0.45])