Make short video factory use Comfy templates

src/routes/money_printer.rs

@@ -14,7 +14,7 @@ use axum::{
 };
 use reqwest::Client;
 use serde::Deserialize;
-use serde_json::{json, Value};
+use serde_json::{json, Map, Value};
 use std::time::Duration;
 
 const TASK_TYPE_MONEY_PRINTER: &str = "money-printer-video";
@@ -585,6 +585,24 @@ fn first_node_by_class(prompt: &serde_json::Map<String, Value>, class_type: &str
         .map(|(id, _)| id.clone())
 }
 
+fn first_existing_input_field(
+    prompt: &serde_json::Map<String, Value>,
+    node_id: &str,
+    fields: &[&str],
+    fallback: &str,
+) -> String {
+    let inputs = prompt
+        .get(node_id)
+        .and_then(|node| node.get("inputs"))
+        .and_then(Value::as_object);
+    fields
+        .iter()
+        .find(|field| inputs.is_some_and(|inputs| inputs.contains_key(**field)))
+        .copied()
+        .unwrap_or(fallback)
+        .to_string()
+}
+
 fn prompt_contains_class_type(prompt: &Value, class_type: &str) -> bool {
     match prompt {
         Value::Object(object) => {
@@ -616,10 +634,14 @@ fn infer_comfy_mapping(prompt: &Value) -> Value {
         .or_else(|| first_node_by_class(object, "KSampler"))
         .or_else(|| first_node_by_class(object, "EmptyLatentImage"))
         .unwrap_or_default();
+    let prompt_field = first_existing_input_field(object, &prompt_node, &["prompt", "positive", "text"], "text");
+    let ratio_field = first_existing_input_field(object, &video_node, &["ratio", "model.ratio", "aspect_ratio"], "ratio");
+    let resolution_field = first_existing_input_field(object, &video_node, &["model.resolution", "resolution", "size"], "resolution");
+    let duration_field = first_existing_input_field(object, &video_node, &["model.duration", "duration", "length"], "duration");
     json!({
         "prompt": {
             "nodeId": prompt_node,
-            "field": "text"
+            "field": prompt_field
         },
         "negativePrompt": {
             "nodeId": negative_node,
@@ -627,15 +649,15 @@ fn infer_comfy_mapping(prompt: &Value) -> Value {
         },
         "ratio": {
             "nodeId": video_node,
-            "field": "ratio"
+            "field": ratio_field
         },
         "resolution": {
             "nodeId": video_node,
-            "field": "resolution"
+            "field": resolution_field
         },
         "duration": {
             "nodeId": video_node,
-            "field": "duration"
+            "field": duration_field
         }
     })
 }
@@ -744,6 +766,47 @@ fn apply_node_input(prompt: &mut Value, node_id: &str, field: &str, value: Value
     Ok(())
 }
 
+fn node_has_input(prompt: &Value, node_id: &str, field: &str) -> bool {
+    prompt
+        .get(node_id)
+        .and_then(|node| node.get("inputs"))
+        .and_then(Value::as_object)
+        .is_some_and(|inputs| inputs.contains_key(field))
+}
+
+fn apply_optional_node_input(
+    prompt: &mut Value,
+    mapping: &Value,
+    mapping_keys: &[&str],
+    default_field: &str,
+    value: Value,
+) {
+    let Some(node_id) = mapping_node_id(mapping, mapping_keys) else {
+        return;
+    };
+    let field = mapping_field(mapping, mapping_keys, default_field);
+    if !node_has_input(prompt, &node_id, &field) {
+        return;
+    }
+    let _ = apply_node_input(prompt, &node_id, &field, value);
+}
+
+fn value_to_duration_seconds(value: &Value) -> Option<i64> {
+    value
+        .as_i64()
+        .or_else(|| value.as_str().and_then(|text| text.trim().parse::<i64>().ok()))
+        .map(|duration| duration.clamp(2, 120))
+}
+
+fn normalize_resolution_for_comfy(value: &Value) -> Value {
+    let text = value.as_str().unwrap_or_default().trim().to_ascii_lowercase();
+    match text.as_str() {
+        "720p" | "720" => json!("720P"),
+        "1080p" | "1080" => json!("1080P"),
+        _ => value.clone(),
+    }
+}
+
 fn inject_comfy_video_inputs(prompt: &mut Value, mapping: &Value, body: &Value, plan: &Value) -> AppResult<String> {
     let video_prompt = string_from(body, &["videoPrompt", "video_prompt"]).unwrap_or_else(|| {
         plan.get("video_prompt")
@@ -773,18 +836,117 @@ fn inject_comfy_video_inputs(prompt: &mut Value, mapping: &Value, body: &Value,
             json!(negative),
         )?;
     }
-    for (keys, request_keys, default_field) in [
+    if let Some(value) = ["ratio", "aspectRatio", "aspect_ratio"].iter().find_map(|key| body.get(*key).cloned()) {
-        (&["ratio", "ratioNode"][..], &["ratio", "aspectRatio", "aspect_ratio"][..], "ratio"),
+        apply_optional_node_input(prompt, mapping, &["ratio", "ratioNode"], "ratio", value);
-        (&["resolution", "resolutionNode"][..], &["resolution", "size"][..], "resolution"),
+    }
-        (&["duration", "durationNode"][..], &["duration"][..], "duration"),
+    if let Some(value) = ["resolution", "size"].iter().find_map(|key| body.get(*key).cloned()) {
-    ] {
+        apply_optional_node_input(
-        if let Some(node_id) = mapping_node_id(mapping, keys) {
+            prompt,
-            if let Some(value) = request_keys.iter().find_map(|key| body.get(*key).cloned()) {
+            mapping,
-                apply_node_input(prompt, &node_id, &mapping_field(mapping, keys, default_field), value)?;
+            &["resolution", "resolutionNode"],
-            }
+            "resolution",
+            normalize_resolution_for_comfy(&value),
+        );
+    }
+    if let Some(value) = ["duration"].iter().find_map(|key| body.get(*key)).and_then(value_to_duration_seconds) {
+        apply_optional_node_input(prompt, mapping, &["duration", "durationNode"], "duration", json!(value));
+    }
+    repair_money_printer_comfy_prompt(prompt);
+    Ok(video_prompt)
+}
+
+fn repair_money_printer_comfy_prompt(prompt: &mut Value) {
+    repair_resize_image_mask_nodes(prompt);
+    repair_text_generate_ltx2_prompt_nodes(prompt);
+    repair_wan2_api_nodes(prompt);
+}
+
+fn repair_resize_image_mask_nodes(prompt: &mut Value) {
+    let Some(nodes) = prompt.as_object_mut() else {
+        return;
+    };
+    for (_node_id, node) in nodes.iter_mut() {
+        if node.get("class_type").and_then(Value::as_str) != Some("ResizeImageMaskNode") {
+            continue;
+        }
+        let Some(inputs) = node.get_mut("inputs").and_then(Value::as_object_mut) else {
+            continue;
+        };
+        if inputs
+            .get("resize_type.crop")
+            .is_some_and(|value| !matches!(value.as_str(), Some("disabled" | "center")))
+        {
+            inputs.insert("resize_type.crop".to_string(), Value::String("center".to_string()));
         }
     }
-    Ok(video_prompt)
+}
+
+fn repair_text_generate_ltx2_prompt_nodes(prompt: &mut Value) {
+    let Some(nodes) = prompt.as_object_mut() else {
+        return;
+    };
+    for (_node_id, node) in nodes.iter_mut() {
+        if node.get("class_type").and_then(Value::as_str) != Some("TextGenerateLTX2Prompt") {
+            continue;
+        }
+        let Some(inputs) = node.get_mut("inputs").and_then(Value::as_object_mut) else {
+            continue;
+        };
+        if inputs
+            .get("sampling_mode.repetition_penalty")
+            .and_then(Value::as_f64)
+            .is_some_and(|value| !(0.0..=5.0).contains(&value))
+        {
+            inputs.insert("sampling_mode".to_string(), Value::String("on".to_string()));
+            inputs.insert("sampling_mode.temperature".to_string(), json!(0.7));
+            inputs.insert("sampling_mode.top_k".to_string(), json!(64));
+            inputs.insert("sampling_mode.top_p".to_string(), json!(0.95));
+            inputs.insert("sampling_mode.min_p".to_string(), json!(0.05));
+            inputs.insert("sampling_mode.repetition_penalty".to_string(), json!(1.05));
+            inputs.entry("sampling_mode.seed".to_string()).or_insert_with(|| json!(0));
+            inputs.insert("sampling_mode.presence_penalty".to_string(), json!(0.0));
+        }
+    }
+}
+
+fn repair_wan2_api_nodes(prompt: &mut Value) {
+    let Some(nodes) = prompt.as_object_mut() else {
+        return;
+    };
+    for (_node_id, node) in nodes.iter_mut() {
+        let class_type = node.get("class_type").and_then(Value::as_str).unwrap_or_default();
+        if !class_type.starts_with("Wan2") || !class_type.ends_with("Api") {
+            continue;
+        }
+        let Some(inputs) = node.get_mut("inputs").and_then(Value::as_object_mut) else {
+            continue;
+        };
+        repair_wan2_model_group(inputs);
+    }
+}
+
+fn repair_wan2_model_group(inputs: &mut Map<String, Value>) {
+    let model_value = inputs.get("model").cloned().unwrap_or(Value::Null);
+    if let Some(items) = model_value.as_array() {
+        if let Some(model) = items.first().cloned() {
+            inputs.insert("model".to_string(), model);
+        }
+        if let Some(resolution) = items.get(1).cloned() {
+            inputs.insert("model.resolution".to_string(), normalize_resolution_for_comfy(&resolution));
+        }
+        if let Some(duration) = items.get(2).and_then(value_to_duration_seconds) {
+            inputs.insert("model.duration".to_string(), json!(duration));
+        }
+    }
+    if inputs
+        .get("model.resolution")
+        .is_some_and(|value| !matches!(value.as_str(), Some("720P" | "1080P")))
+    {
+        inputs.insert("model.resolution".to_string(), json!("1080P"));
+    }
+    if let Some(duration) = inputs.get("model.duration").and_then(value_to_duration_seconds) {
+        inputs.insert("model.duration".to_string(), json!(duration));
+    }
 }
 
 async fn submit_comfy_prompt(state: &AppState, prompt: Value) -> AppResult<(String, Value)> {

src/web/pages/MoneyPrinter.tsx

@@ -55,6 +55,11 @@ function mappingFromGraph(graph: any) {
 
 function inferMappingFromPrompt(prompt: any) {
   const nodes = prompt && typeof prompt === 'object' && !Array.isArray(prompt) ? Object.entries(prompt as Record<string, any>) : [];
+  const inputKeys = (nodeId: string) => Object.keys(prompt?.[String(nodeId)]?.inputs || {});
+  const firstInputField = (nodeId: string, fields: string[], fallback: string) => {
+    const keys = inputKeys(nodeId);
+    return fields.find((field) => keys.includes(field)) || fallback;
+  };
   const find = (classWords: string[], inputWords: string[] = []) => {
     return nodes.find(([, node]) => {
       const classType = String(node?.class_type || node?.type || '').toLowerCase();
@@ -66,11 +71,11 @@ function inferMappingFromPrompt(prompt: any) {
   const negativeNode = find(['negative'], ['negative']);
   const videoNode = find(['wan', 'ltx', 'video'], ['duration', 'ratio', 'resolution']);
   return {
-    prompt: { nodeId: String(promptNode), field: prompt?.[String(promptNode)]?.inputs?.prompt !== undefined ? 'prompt' : 'text' },
+    prompt: { nodeId: String(promptNode), field: firstInputField(String(promptNode), ['prompt', 'positive', 'text', 'positive_prompt'], 'text') },
     negativePrompt: { nodeId: String(negativeNode), field: 'text' },
-    ratio: { nodeId: String(videoNode), field: 'ratio' },
+    ratio: { nodeId: String(videoNode), field: firstInputField(String(videoNode), ['ratio', 'model.ratio', 'aspect_ratio'], 'ratio') },
-    resolution: { nodeId: String(videoNode), field: 'resolution' },
+    resolution: { nodeId: String(videoNode), field: firstInputField(String(videoNode), ['model.resolution', 'resolution', 'size'], 'resolution') },
-    duration: { nodeId: String(videoNode), field: 'duration' },
+    duration: { nodeId: String(videoNode), field: firstInputField(String(videoNode), ['model.duration', 'duration', 'length'], 'duration') },
   };
 }
 
@@ -192,7 +197,6 @@ function workflowLooksVideoCapable(workflow: WorkflowSummary) {
 function moneyPrinterWorkflows(items: WorkflowSummary[]) {
   const seen = new Set<string>();
   return items
-    .filter((workflow) => workflowLooksVideoCapable(workflow))
     .filter((workflow) => {
       const key = workflow.publicId || String(workflow.id);
       if (seen.has(key)) return false;
@@ -200,6 +204,9 @@ function moneyPrinterWorkflows(items: WorkflowSummary[]) {
       return true;
     })
     .sort((a, b) => {
+      const aVideo = workflowLooksVideoCapable(a);
+      const bVideo = workflowLooksVideoCapable(b);
+      if (aVideo !== bVideo) return aVideo ? -1 : 1;
       const aTemplate = a.isTemplate || a.visibility === 'template' || a.visibility === 'public';
       const bTemplate = b.isTemplate || b.visibility === 'template' || b.visibility === 'public';
       if (aTemplate !== bTemplate) return aTemplate ? -1 : 1;
@@ -207,6 +214,16 @@ function moneyPrinterWorkflows(items: WorkflowSummary[]) {
     });
 }
 
+function parseJsonInput(value: string, label: string) {
+  const trimmed = value.trim();
+  if (!trimmed) return undefined;
+  try {
+    return JSON.parse(trimmed);
+  } catch {
+    throw new Error(`${label} 不是有效 JSON`);
+  }
+}
+
 export default function MoneyPrinter() {
   const location = useLocation();
   const audience: SessionAudience = location.pathname.startsWith('/user/') ? 'user' : 'admin';
@@ -215,7 +232,7 @@ export default function MoneyPrinter() {
   const [topic, setTopic] = useState('帮我生成一个 30 秒的 AI 应用平台 BossAI 商品介绍短视频');
   const [scriptModel, setScriptModel] = useState('gpt-5.4');
   const [videoModel, setVideoModel] = useState('seedance-1-0-lite-t2v');
-  const [videoBackend, setVideoBackend] = useState<'boosapi' | 'comfy'>('boosapi');
+  const [videoBackend, setVideoBackend] = useState<'boosapi' | 'comfy'>('comfy');
   const [style, setStyle] = useState(STYLE_PRESETS[0]);
   const [audienceText, setAudienceText] = useState('AI 工具用户、内容创作者、企业运营');
   const [duration, setDuration] = useState(30);
@@ -226,7 +243,7 @@ export default function MoneyPrinter() {
   const [comfyNodeMapping, setComfyNodeMapping] = useState('{\n  "prompt": { "nodeId": "", "field": "prompt" },\n  "negativePrompt": { "nodeId": "", "field": "text" },\n  "ratio": { "nodeId": "", "field": "ratio" },\n  "resolution": { "nodeId": "", "field": "resolution" },\n  "duration": { "nodeId": "", "field": "duration" }\n}');
   const [workflows, setWorkflows] = useState<WorkflowSummary[]>([]);
   const [templateId, setTemplateId] = useState('');
-  const [useAdvancedComfy, setUseAdvancedComfy] = useState(false);
+  const [showAdvancedComfy, setShowAdvancedComfy] = useState(false);
   const [jobs, setJobs] = useState<MoneyPrinterJob[]>([]);
   const [submitting, setSubmitting] = useState(false);
   const [error, setError] = useState('');
@@ -254,9 +271,10 @@ export default function MoneyPrinter() {
   }, [loadJobs, loadWorkflows]);
 
   const selectedWorkflow = workflows.find((item) => item.publicId === templateId) || null;
+  const canSubmit = Boolean(topic.trim()) && !submitting && (videoBackend !== 'comfy' || Boolean(templateId));
 
   useEffect(() => {
-    if (!selectedWorkflow || videoBackend !== 'comfy' || !useAdvancedComfy) return;
+    if (!selectedWorkflow || videoBackend !== 'comfy') return;
     const prompt = workflowPromptFromGraph(selectedWorkflow.graph);
     if (prompt) setComfyPrompt(JSON.stringify(prompt, null, 2));
     const mapping = mappingFromGraph(selectedWorkflow.graph);
@@ -292,8 +310,8 @@ export default function MoneyPrinter() {
         language: '中文',
         ...(videoBackend === 'comfy' ? {
           templateId: templateId || undefined,
-          comfyPrompt: comfyPrompt.trim() ? JSON.parse(comfyPrompt) : undefined,
+          comfyPrompt: parseJsonInput(comfyPrompt, 'Comfy prompt JSON'),
-          comfyNodeMapping: comfyNodeMapping.trim() ? JSON.parse(comfyNodeMapping) : undefined,
+          comfyNodeMapping: parseJsonInput(comfyNodeMapping, '节点映射 JSON'),
         } : {}),
       });
       setJobs((prev) => [payload.job, ...prev.filter((item) => item.id !== payload.job.id)]);
@@ -329,7 +347,7 @@ export default function MoneyPrinter() {
     setComfyPrompt(request.comfyPrompt ? JSON.stringify(request.comfyPrompt, null, 2) : comfyPrompt);
     setComfyNodeMapping(request.comfyNodeMapping ? JSON.stringify(request.comfyNodeMapping, null, 2) : comfyNodeMapping);
     setTemplateId(request.templateId || templateId);
-    setUseAdvancedComfy(Boolean(request.templateId));
+    setShowAdvancedComfy(Boolean(request.comfyPrompt || request.comfyNodeMapping));
   };
 
   return (
@@ -341,7 +359,7 @@ export default function MoneyPrinter() {
         </div>
         <div className="page-actions">
           <button type="button" className="btn btn-ghost" onClick={() => void loadJobs()}>刷新</button>
-          <button type="button" className="btn btn-primary" onClick={submit} disabled={submitting || !topic.trim()}>
+          <button type="button" className="btn btn-primary" onClick={submit} disabled={!canSubmit}>
             {submitting ? '提交中...' : '生成短视频'}
           </button>
         </div>
@@ -353,7 +371,7 @@ export default function MoneyPrinter() {
         <section className="money-printer-panel">
           <div className="money-printer-hero">
             <span>MoneyPrinterTurbo inside</span>
-            <strong>一句主题，自动拆成脚本、分镜、视频提示词和成片任务。</strong>
+            <strong>一句主题，自动拆成脚本、分镜、视频提示词，再提交到 Comfy API 生成成片。</strong>
           </div>
 
           <label className="money-printer-field">
@@ -375,7 +393,12 @@ export default function MoneyPrinter() {
             </label>
             <label className="money-printer-field">
               <span>视频模型</span>
-              <input value={videoModel} onChange={(event) => setVideoModel(event.target.value)} />
+              <input
+                value={videoModel}
+                onChange={(event) => setVideoModel(event.target.value)}
+                disabled={videoBackend === 'comfy'}
+                title={videoBackend === 'comfy' ? 'Comfy API 模式使用所选工作流模板里的模型' : undefined}
+              />
             </label>
             <label className="money-printer-field">
               <span>时长</span>
@@ -430,7 +453,7 @@ export default function MoneyPrinter() {
                   <option value="">请选择模板</option>
                   {workflows.map((workflow) => (
                     <option key={workflow.publicId} value={workflow.publicId}>
-                      {workflowOptionLabel(workflow)}
+                      {workflowLooksVideoCapable(workflow) ? '视频 · ' : '模板 · '}{workflowOptionLabel(workflow)}
                     </option>
                   ))}
                 </select>
@@ -438,17 +461,32 @@ export default function MoneyPrinter() {
               {selectedWorkflow ? (
                 <div className="money-printer-template-meta">
                   <strong>{selectedWorkflow.name}</strong>
-                  <span>{mappingFromGraph(selectedWorkflow.graph) ? '已带节点映射' : '这个模板没有内置节点映射，系统会尝试自动推断。'}</span>
+                  <span>
+                    {workflowLooksVideoCapable(selectedWorkflow) ? '已识别为视频工作流' : '未识别视频关键词，请确认这是 Comfy 视频模板'}
+                    {' · '}
+                    {mappingFromGraph(selectedWorkflow.graph) ? '已带节点映射' : '系统会按节点结构自动推断提示词和参数映射'}
+                  </span>
                 </div>
               ) : null}
-              <label className="money-printer-field">
+              <button
-                <span>Comfy prompt JSON</span>
+                type="button"
-                <textarea value={comfyPrompt} onChange={(event) => setComfyPrompt(event.target.value)} rows={7} placeholder='{"12":{"class_type":"...","inputs":{...}}}' />
+                className="btn btn-ghost money-printer-advanced-toggle"
-              </label>
+                onClick={() => setShowAdvancedComfy((prev) => !prev)}
-              <label className="money-printer-field">
+              >
-                <span>节点映射 JSON</span>
+                {showAdvancedComfy ? '收起模板调试' : '展开模板调试'}
-                <textarea value={comfyNodeMapping} onChange={(event) => setComfyNodeMapping(event.target.value)} rows={7} />
+              </button>
-              </label>
+              <div className={`anim-collapse ${showAdvancedComfy ? 'is-open' : ''}`.trim()}>
+                <div className="anim-collapse-inner">
+                  <label className="money-printer-field">
+                    <span>Comfy prompt JSON</span>
+                    <textarea value={comfyPrompt} onChange={(event) => setComfyPrompt(event.target.value)} rows={7} placeholder='{"12":{"class_type":"...","inputs":{...}}}' />
+                  </label>
+                  <label className="money-printer-field">
+                    <span>节点映射 JSON</span>
+                    <textarea value={comfyNodeMapping} onChange={(event) => setComfyNodeMapping(event.target.value)} rows={7} />
+                  </label>
+                </div>
+              </div>
             </div>
           )}
         </section>