diff --git a/webui/configs/model_params.json b/webui/configs/model_params.json index 7b3d71bbc..d24ff82c1 100644 --- a/webui/configs/model_params.json +++ b/webui/configs/model_params.json @@ -1,6 +1,6 @@ { "echo_tts": [ - {"name": "num_inference_steps", "type": "slider", "label": "num_inference_steps", "label_en": "Sampling steps", "default": 40, "minimum": 8, "maximum": 40, "step": 1, "precision": 0, "info": "Euler sampler steps."}, + {"name": "num_inference_steps", "type": "slider", "label": "num_inference_steps", "label_en": "Sampling steps", "default": 40, "minimum": 8, "maximum": 120, "step": 1, "precision": 0, "info": "Euler sampler steps.", "info_en": "Euler sampler steps. No upper bound in the model; higher is slower and steadier."}, {"name": "text_guidance_scale", "type": "slider", "label": "text_guidance_scale", "label_en": "Text guidance", "default": 3.0, "minimum": 0.0, "maximum": 10.0, "step": 0.1}, {"name": "speaker_guidance_scale", "type": "slider", "label": "speaker_guidance_scale", "label_en": "Speaker guidance", "default": 8.0, "minimum": 0.0, "maximum": 15.0, "step": 0.1}, {"name": "truncation_factor", "type": "slider", "label": "truncation_factor", "label_en": "Noise truncation", "default": 0.8, "minimum": 0.0, "maximum": 1.0, "step": 0.05}, @@ -11,7 +11,7 @@ "_comment": "WebUI TTS 高级参数控件配置:按模型 family 动态生成控件(gr.render)。每项字段:name=选项键(随请求 options 透传给模型);type=slider|number|bool|text|choice;label/info=显示文案;default=默认值(应等于模型默认,已按 src/models//*.cpp 校对);minimum/maximum/step=数值范围;precision=0 表示整数;choices=下拉候选。规则:只有被用户改动过的控件值才会随请求发送;seed/max_tokens 已有专用输入框,勿在此重复;参考文本用『参考文本』框(reference_text);文件路径/parity 类参数(如 *_noise_file)未纳入,可用『其它参数(JSON)』兜底框传。", "qwen3_tts": [ - {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.9, "minimum": 0.0, "maximum": 2.0, "step": 0.05}, + {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.9, "minimum": 0.05, "maximum": 2.0, "step": 0.05, "info_en": "Must be above 0 while do_sample is on."}, {"name": "top_k", "type": "number", "label": "top_k", "default": 50, "minimum": 0, "step": 1, "precision": 0}, {"name": "top_p", "type": "slider", "label": "top_p", "default": 1.0, "minimum": 0.0, "maximum": 1.0, "step": 0.01}, {"name": "repetition_penalty", "type": "slider", "label": "repetition_penalty", "default": 1.05, "minimum": 1.0, "maximum": 2.0, "step": 0.01}, @@ -111,16 +111,16 @@ {"name": "guidance_scale", "type": "slider", "label": "guidance_scale", "default": 2.5, "minimum": 0.0, "maximum": 8.0, "step": 0.1}, {"name": "spatio_temporal_guidance_scale", "type": "slider", "label": "spatio_temporal_guidance_scale", "default": 1.5, "minimum": 0.0, "maximum": 5.0, "step": 0.1}, {"name": "duration_scale", "type": "slider", "label": "duration_scale(自动估时倍率)", "label_en": "duration_scale", "default": 1.1, "minimum": 0.5, "maximum": 2.0, "step": 0.05}, - {"name": "reference_duration_sec", "type": "number", "label": "reference_duration_sec(参考音频裁剪/重复秒数)", "label_en": "reference_duration_sec", "default": 10.0, "minimum": 0.0, "step": 0.5}, + {"name": "reference_duration_sec", "type": "number", "label": "reference_duration_sec(参考音频裁剪/重复秒数)", "label_en": "reference_duration_sec", "default": 10.0, "minimum": 0.5, "step": 0.5, "info_en": "Must be positive."}, {"name": "guidance_rescale", "type": "text", "label": "guidance_rescale", "default": "auto", "placeholder": "auto 或数值"}, - {"name": "audio_chunk_threshold_sec", "type": "number", "label": "audio_chunk_threshold_sec(长文本阈值)", "label_en": "audio_chunk_threshold_sec", "default": 45.0, "minimum": 0.0, "step": 1.0}, - {"name": "audio_chunk_duration_sec", "type": "number", "label": "audio_chunk_duration_sec(长文本分段目标时长)", "label_en": "audio_chunk_duration_sec", "default": 37.0, "minimum": 0.0, "step": 1.0}, + {"name": "audio_chunk_threshold_sec", "type": "number", "label": "audio_chunk_threshold_sec(长文本阈值)", "label_en": "audio_chunk_threshold_sec", "default": 45.0, "minimum": 1.0, "step": 1.0, "info_en": "Must be positive."}, + {"name": "audio_chunk_duration_sec", "type": "number", "label": "audio_chunk_duration_sec(长文本分段目标时长)", "label_en": "audio_chunk_duration_sec", "default": 37.0, "minimum": 1.0, "step": 1.0, "info_en": "Must be positive."}, {"name": "cross_fade_duration_sec", "type": "number", "label": "cross_fade_duration_sec(分段交叉淡化)", "label_en": "cross_fade_duration_sec", "default": 0.05, "minimum": 0.0, "step": 0.01} ], "confucius4_tts": [ - {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.8, "minimum": 0.0, "maximum": 2.0, "step": 0.05}, - {"name": "top_p", "type": "slider", "label": "top_p", "default": 0.8, "minimum": 0.0, "maximum": 1.0, "step": 0.01}, + {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.8, "minimum": 0.05, "maximum": 2.0, "step": 0.05, "info_en": "Must be positive."}, + {"name": "top_p", "type": "slider", "label": "top_p", "default": 0.8, "minimum": 0.01, "maximum": 1.0, "step": 0.01, "info_en": "Must be within (0, 1]."}, {"name": "top_k", "type": "number", "label": "top_k", "default": 30, "minimum": 1, "step": 1, "precision": 0}, {"name": "num_beams", "type": "number", "label": "num_beams", "default": 3, "minimum": 1, "step": 1, "precision": 0}, {"name": "repetition_penalty", "type": "slider", "label": "repetition_penalty", "default": 10.0, "minimum": 0.0, "maximum": 20.0, "step": 0.1}, @@ -206,8 +206,8 @@ "personaplex": [ {"name": "voice_id", "type": "choice", "label": "voice_id(打包音色)", "label_en": "voice_id (packaged voice)", "default": "NATF2", "choices": ["NATF0", "NATF1", "NATF2", "NATF3", "NATM0", "NATM1", "NATM2", "NATM3", "VARF0", "VARF1", "VARF2", "VARF3", "VARF4", "VARM0", "VARM1", "VARM2", "VARM3", "VARM4"]}, {"name": "system_prompt", "type": "text", "label": "system_prompt", "default": "", "placeholder": "Leave blank to use the text box as the system prompt."}, - {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.8, "minimum": 0.0, "maximum": 2.0, "step": 0.05}, - {"name": "text_temperature", "type": "slider", "label": "text_temperature", "default": 0.8, "minimum": 0.0, "maximum": 2.0, "step": 0.05}, + {"name": "temperature", "type": "slider", "label": "temperature", "default": 0.8, "minimum": 0.05, "maximum": 2.0, "step": 0.05, "info_en": "Must be above 0 while do_sample is on."}, + {"name": "text_temperature", "type": "number", "label": "text_temperature", "label_en": "text_temperature (blank follows temperature)", "minimum": 0.05, "maximum": 2.0, "step": 0.05, "info_en": "Blank makes the text head follow temperature. Must be above 0 while do_sample is on."}, {"name": "top_k", "type": "number", "label": "top_k", "default": 250, "minimum": 0, "step": 1, "precision": 0}, {"name": "text_top_k", "type": "number", "label": "text_top_k", "default": 250, "minimum": 0, "step": 1, "precision": 0}, {"name": "do_sample", "type": "bool", "label": "do_sample", "default": true} @@ -217,9 +217,9 @@ {"name": "route", "type": "choice", "label": "route(任务路线)", "default": "", "choices": ["", "style_preserved_vc", "style_converted_vc", "style_preserved_svc", "style_converted_svc", "singing_style_conversion", "editing"], "info": "留空=按任务默认;详见 webui/README.md"}, {"name": "num_inference_steps", "type": "number", "label": "num_inference_steps", "default": 32, "minimum": 1, "step": 1, "precision": 0, "info": "流匹配步数"}, {"name": "use_pitch_shift", "type": "choice", "label": "use_pitch_shift(自动音高对齐)", "default": "", "choices": ["", "true", "false"], "info": "留空=按路线默认"}, - {"name": "temperature", "type": "slider", "label": "temperature(AR 路线用)", "default": 0.7, "minimum": 0.0, "maximum": 2.0, "step": 0.05, "info": "默认取自模型 generation_config.json"}, + {"name": "temperature", "type": "number", "label": "temperature(AR 路线用)", "label_en": "temperature (AR paths)", "minimum": 0.05, "maximum": 2.0, "step": 0.05, "info": "留空=取自模型 generation_config.json(回退 1.0)", "info_en": "Blank uses the checkpoint's generation_config.json (C++ fallback 1.0). Must be above 0."}, {"name": "top_k", "type": "number", "label": "top_k(AR 路线用)", "default": 20, "minimum": 0, "step": 1, "precision": 0, "info": "默认取自模型 generation_config.json"}, - {"name": "top_p", "type": "slider", "label": "top_p(AR 路线用)", "default": 0.8, "minimum": 0.0, "maximum": 1.0, "step": 0.01} + {"name": "top_p", "type": "number", "label": "top_p(AR 路线用)", "label_en": "top_p (AR paths)", "minimum": 0.0, "maximum": 1.0, "step": 0.01, "info": "留空=取自模型 generation_config.json(回退 0.8)", "info_en": "Blank uses the checkpoint's generation_config.json (C++ fallback 0.8)."} ], "heartmula": [ @@ -273,9 +273,9 @@ ], "controlfoley": [ - {"name": "duration_sec", "type": "number", "label": "duration_sec", "default": 8.0, "minimum": 0.1, "step": 0.5}, + {"name": "duration_sec", "type": "number", "label": "duration_sec", "default": 8.0, "minimum": 0.64, "maximum": 60.0, "step": 0.1, "info_en": "Shorter than 0.64 s produces an empty sync sequence and fails."}, {"name": "num_inference_steps", "type": "number", "label": "num_inference_steps", "default": 25, "minimum": 1, "step": 1, "precision": 0}, - {"name": "guidance_scale", "type": "slider", "label": "guidance_scale", "default": 4.5, "minimum": 0.0, "maximum": 10.0, "step": 0.1}, + {"name": "guidance_scale", "type": "slider", "label": "guidance_scale", "default": 4.5, "minimum": 0.1, "maximum": 10.0, "step": 0.1, "info_en": "Must be positive."}, {"name": "negative_prompt", "type": "text", "label": "negative_prompt", "default": "", "placeholder": "optional negative prompt"}, {"name": "mask_away_clip", "type": "bool", "label": "mask_away_clip", "default": false} ], @@ -284,7 +284,7 @@ {"name": "duration_sec", "type": "number", "label": "duration_sec", "default": 20.0, "minimum": 0.1, "step": 0.5}, {"name": "guidance_scale", "type": "slider", "label": "guidance_scale", "default": 2.0, "minimum": 0.0, "maximum": 10.0, "step": 0.1}, {"name": "stop_threshold", "type": "slider", "label": "stop_threshold", "default": 0.5, "minimum": 0.0, "maximum": 1.0, "step": 0.05}, - {"name": "min_stop_step", "type": "number", "label": "min_stop_step", "default": 5, "minimum": 0, "step": 1, "precision": 0} + {"name": "min_stop_step", "type": "number", "label": "min_stop_step", "default": 5, "minimum": 1, "step": 1, "precision": 0, "info_en": "Must be positive."} ], "fireredtts3-base": [