优化完善替换角色

2026-09-01 18:29:36 +08:00
parent 0603a85d83
commit 1825543873
4 changed files with 64 additions and 121 deletions
@@ -729,19 +729,14 @@ class SubmitVideoReplaceTests(TestCase):
        """
        from apps.ai.video_replace import build_character_submit_prompt

-        with_voice = build_character_submit_prompt("男性，30 岁左右，中低音域", has_voice_ref=True)
-        self.assertIn("@参考角色", with_voice)
-        self.assertIn("@参考音频", with_voice)
-        self.assertIn("@参考视频", with_voice)
-        # brief 不进提示词
-        self.assertNotIn("30 岁", with_voice)
+        prompt = build_character_submit_prompt()
+        self.assertIn("@参考角色", prompt)
+        self.assertIn("@参考视频", prompt)
+        # 只换画面,不碰声音
+        self.assertNotIn("@参考音频", prompt)
+        self.assertNotIn("音色", prompt)
        # 一句话:不该有分段
-        self.assertNotIn("\n", with_voice)
-
-        # 没有音色样音时不能点名一个不存在的 @参考音频
-        no_voice = build_character_submit_prompt("短发", has_voice_ref=False)
-        self.assertIn("@参考角色", no_voice)
-        self.assertNotIn("@参考音频", no_voice)
+        self.assertNotIn("\n", prompt)

    def test_character_brief_lands_in_submitted_prompt(self):
        portrait = _asset(self.team, self.user, name="模特.png", preview="http://tos/model.png")
@@ -753,25 +748,21 @@ class SubmitVideoReplaceTests(TestCase):
        self.assertEqual(task.request_payload["subject_brief"], "男性，稳重，中低音域")
        self.assertNotIn("声音重建规则", task.request_payload["prompt"])

-    def test_character_gender_attaches_platform_voice_reference(self):
-        task = self._submit_character(subject_brief="男性，稳重，中低音域")
-        self.assertEqual(task.request_payload["voice_gender"], "male")
-        self.assertEqual(task.request_payload["voice_strategy"], "native_audio_reference")
-        # 平台预置音色样音要跟着一起发给 Seedance,并在提示词里被点名
-        audio = [r for r in task.request_payload["references"] if r.get("type") == "audio"]
-        self.assertEqual(len(audio), 1)
-        self.assertEqual(audio[0]["label"], "参考音频")
-        self.assertEqual(audio[0]["role"], "reference_audio")
-        self.assertNotIn("asset_id", audio[0], "平台音频不能带 asset_id，否则会被送去真人素材审核卡死")
-        self.assertIn("@参考音频", task.request_payload["prompt"])
-        self.assertFalse(task.request_payload["digest_pending"], "音频参考路线不需要拆镜")
-
-        # 没写明性别就不该带音频参考,也不该出现声音段落
-        plain = self._submit_character(subject_brief="短发，深色衬衫")
-        self.assertEqual(plain.request_payload["voice_strategy"], "preserve_original")
-        self.assertNotIn("reference_audio", [r.get("role") for r in plain.request_payload["references"]])
-        self.assertNotIn("@参考音频", plain.request_payload["prompt"])
+    def test_character_keeps_original_audio(self):
+        """角色复刻只换画面,声音保留原声。

+        历史:试过四条让 Seedance 换声的路(提示词声线规则、音频参考走直链、走 asset://、
+        平台预置样音),最好也只做到「中途才切到目标音色」。产品上已收敛为同性别替换 +
+        保留原声,前端明确告知用户 —— 别再往提示词或 references 里加音频。
+        """
+        task = self._submit_character(subject_brief="30 岁男性，短发")
+        payload = task.request_payload
+        self.assertNotIn("reference_audio", [r.get("role") for r in payload["references"]])
+        self.assertNotIn("audio", [r.get("type") for r in payload["references"]])
+        self.assertNotIn("@参考音频", payload["prompt"])
+        self.assertNotIn("音色", payload["prompt"])
+        self.assertNotIn("voice_gender", payload)
+        self.assertNotIn("voice_strategy", payload)

    def test_character_library_and_temp_are_exclusive(self):
        portrait = _asset(self.team, self.user, name="模特.png", preview="http://tos/model.png")
@@ -83,54 +83,25 @@ PRODUCT_PROMPT = (
    "按下面的商品语义重构分镜生成视频。@目标商品只用于锁定外观、材质与包装。"
    "严格遵守每镜的新商品动作、场景与口播；不要恢复参考片原商品、原动作、原卖点或原台词。"
 )
-# 角色复刻提示词:极简单句。之前那版把声音规则写成五大段反复强调,实测是「开头原声、
-# 中途才切到目标音色」;长提示词会稀释关键指令,这版改为一句话直给。
-# ⚠️ 句中的 @参考角色 / @参考音频 必须与 references 的 label 完全一致,否则
-# build_content_items 映射不成「图片N / 音频N」,火山那边就是一串没有指代的字面量。
+# 角色复刻提示词:极简单句。
+# ⚠️ 只换画面,不碰声音 —— 成片保留参考视频原声。
+# 历史(别再走一遍):试过让 Seedance 换声的四条路,提示词声线规则、音频参考走 TOS 直链、
+# 走 asset:// 素材库引用、以及平台预置样音,全部只做到「中途才切到目标音色」或完全无效。
+# 编辑模式(参考视频 + duration=-1 + adaptive)对音轨的处理不受这些输入控制。
+# 因此产品上收敛为:只支持同性别角色替换,声音一律保留原声,前端明确告知用户。
 CHARACTER_PROMPT = (
-    "将@参考视频中出现的角色替换为@参考角色，口播、旁白、画外音的音色替换为@参考音频，"
-    "其它部分例如构图、运镜、人物动作、场景、商品、剪辑、画面节奏等保持与原视频一致。"
-)
-
-# 没有音频参考(没写明性别)时,把声音那半句去掉,免得点名一个不存在的 @参考音频。
-CHARACTER_PROMPT_NO_VOICE = (
    "将@参考视频中出现的角色替换为@参考角色，"
    "其它部分例如构图、运镜、人物动作、场景、商品、剪辑、画面节奏等保持与原视频一致。"
 )


 def build_character_submit_prompt(subject_brief: str = "", *, has_voice_ref: bool = False) -> str:
-    """角色复刻提交提示词。就是那一句,不再追加角色设定 ——
-    subject_brief 只用于推导性别(决定用哪条音色样音),不进提示词,保持指令干净。"""
-    return CHARACTER_PROMPT if has_voice_ref else CHARACTER_PROMPT_NO_VOICE
+    """角色复刻提交提示词。就是那一句。

-
-# 平台预置的目标音色样音(公读直链,不走预签名 —— 参考素材要长期可取)。
-VOICE_REFERENCE_LABEL = "参考音频"
-VOICE_REFERENCE_URLS = {
-    # ⚠️ 换音色样音时务必换文件名,不要覆盖旧文件:上传设的是 immutable + 一年强缓存,
-    # 同名覆盖后 CDN / 火山侧可能仍拿到旧音频,查起来很难受。
-    "male": "https://airshelf.tos-s3-cn-shanghai.volces.com/system/video-replace/voice-references/male-v3.mp3",
-    "female": "https://airshelf.tos-s3-cn-shanghai.volces.com/system/video-replace/voice-references/female-bv001.mp3",
-}
-
-
-def build_voice_reference_ref(gender: str) -> dict | None:
-    """平台预置音色样音的 reference 项。读不出性别就没有音频参考。
-
-    ⚠️ 不带 asset_id:带了会被 _ensure_replace_refs_reviewed 送去火山真人素材审核,
-    而那套是给图片/视频用的,平台音频过不去会把任务卡死。
+    参数保留是为了兼容既有调用点;声音已不再由提示词控制,两个参数都不影响输出。
    """
-    url = VOICE_REFERENCE_URLS.get(gender)
-    if not url:
-        return None
-    return {
-        "url": url,
-        "type": "audio",
-        "role": "reference_audio",
-        "label": VOICE_REFERENCE_LABEL,
-        "source": "platform",
-    }
+    return CHARACTER_PROMPT
+

 # 用户在第 3 步填的补充信息(商品名/品类,或角色性别年龄音色)。选填,不填维持原行为。
 MAX_SUBJECT_BRIEF = 500
@@ -140,20 +111,6 @@ def _clean_subject_brief(value) -> str:
    return str(value or "").strip()[:MAX_SUBJECT_BRIEF]


-def _detect_voice_gender(brief: str) -> str:
-    """从用户填的角色信息里读性别。返回 male / female / ""(读不出)。
-
-    男女都提到(例如「把女的换成男的」)时不下断言 —— 分不清哪个是目标。
-    """
-    has_male = "男" in brief
-    has_female = "女" in brief
-    if has_male and not has_female:
-        return "male"
-    if has_female and not has_male:
-        return "female"
-    return ""
-
-
 # 拆解期间的占位提示词。真正的提示词在 worker 拆完后回写。
 DIGEST_PENDING_PROMPT = "正在提炼参考视频的分镜稿…"

@@ -730,7 +687,6 @@ def serialize_video_replace_task(task, *, include_deleted_assets: bool = False)
        "subject_name": payload.get("subject_name") or "",
        "subject_source": payload.get("subject_source") or "",
        "subject_brief": payload.get("subject_brief") or "",
-        "voice_gender": payload.get("voice_gender") or "",
        "product_id": payload.get("product_id") or "",
        "model_id": payload.get("model_id") or "",
        "review_stage": "reviewing" if reviewing else "",
@@ -793,7 +749,6 @@ def submit_video_replace(*, team, user, params: dict):
    # 第 3 步用户填的补充信息。商品:名称/品类/卖点;角色:性别/年龄/音色。选填。
    subject_brief = _clean_subject_brief(params.get("subject_brief"))
    # 角色描述识别到性别时，成片后调用火山音色转换；识别不到即保留原声。
-    voice_gender = _detect_voice_gender(subject_brief) if replace_mode == "character" else ""

    if has_product:
        subject_name, image_refs, subject_source = _product_library_refs(team, product_id)
@@ -841,8 +796,6 @@ def submit_video_replace(*, team, user, params: dict):
        "model_id": str(model_id) if model_id else "",
        "product_facts": product_facts,
        "subject_brief": subject_brief,
-        "voice_gender": voice_gender,
-        "voice_strategy": "native_audio_reference" if voice_gender else "preserve_original",
        # adaptive / -1 只用于火山请求；预估积分仍按参考视频实际时长和上传时检测出的比例计算。
        "billing_duration": max(4, int(round(video_seconds or 4))),
        "billing_aspect_ratio": billing_ratio,
@@ -868,12 +821,9 @@ def submit_video_replace(*, team, user, params: dict):
        if resolve_digest_model_config() is None:
            raise ValueError("视频提炼模型未配置，请联系管理员")
    if replace_mode == "character":
-        # 平台预置的目标音色样音,和参考视频、角色图一起交给 Seedance。
-        voice_ref = build_voice_reference_ref(voice_gender)
        image_refs = [
            _owned_ref(video, kind="video", role="reference_video", label="参考视频"),
            *image_refs,
-            *([voice_ref] if voice_ref else []),
        ]
    # 商品图/角色图仍要过审才能当生成参考。提前送审 → 审核和提炼并行跑,明确不过审的直接 400。
    review_state = _ensure_replace_refs_reviewed(team, image_refs)
@@ -895,7 +845,7 @@ def submit_video_replace(*, team, user, params: dict):
            **base_params,
            "prompt": (
                DIGEST_PENDING_PROMPT if replace_mode == "product"
-                else build_character_submit_prompt(subject_brief, has_voice_ref=bool(voice_gender))
+                else build_character_submit_prompt()
            ),
            "references": image_refs,
            "extra_payload": extra,
@@ -106,10 +106,6 @@ const REPLACE_MODE_COPY = {
    drawerEmpty: "还没有人物，先去模特库添加",
    historyKind: "角色",
    pickToast: "已选择角色",
-    briefStep: "3. 角色信息",
-    briefOptional: "选填",
-    briefPlaceholder: "例：30 岁左右男性，短发，深色衬衫，气质干练利落",
-    briefHint: "写明目标角色是男或女时，系统会自动使用对应的通用声音参考；未写明则保留参考视频原声。",
  },
 } as const;

@@ -1239,22 +1235,30 @@ export function VideoReplacePage({
                </div>
              </div>

-              <div className="video-flow-step">
-                <div className="video-flow-step-head">
-                  <strong>{copy.briefStep}</strong>
-                  <span>{copy.briefOptional}</span>
+              {replaceMode === "product" ? (
+                <div className="video-flow-step">
+                  <div className="video-flow-step-head">
+                    <strong>{REPLACE_MODE_COPY.product.briefStep}</strong>
+                    <span>{REPLACE_MODE_COPY.product.briefOptional}</span>
+                  </div>
+                  <textarea
+                    className="textarea replace-brief-input"
+                    value={subjectBrief}
+                    maxLength={SUBJECT_BRIEF_MAX}
+                    placeholder={REPLACE_MODE_COPY.product.briefPlaceholder}
+                    disabled={generating}
+                    aria-label={REPLACE_MODE_COPY.product.briefStep}
+                    onChange={(event) => updateCurrentForm({ subjectBrief: event.target.value })}
+                  />
+                  <p className="field-hint replace-brief-hint">{REPLACE_MODE_COPY.product.briefHint}</p>
                </div>
-                <textarea
-                  className="textarea replace-brief-input"
-                  value={subjectBrief}
-                  maxLength={SUBJECT_BRIEF_MAX}
-                  placeholder={copy.briefPlaceholder}
-                  disabled={generating}
-                  aria-label={copy.briefStep}
-                  onChange={(event) => updateCurrentForm({ subjectBrief: event.target.value })}
-                />
-                <p className="field-hint replace-brief-hint">{copy.briefHint}</p>
-              </div>
+              ) : (
+                <div className="tip replace-limit-tip" role="note">
+                  <strong>请替换为同性别的角色</strong>
+                  女性角色请换成女性、男性换成男性。成片会保留参考视频的原声，
+                  所以跨性别替换会出现「画面是男生、声音还是女生」的情况。
+                </div>
+              )}

              <div className="video-flow-actions replace-generate-action">
                <button
@@ -204,19 +204,17 @@
 }

 /* 第 3 步补充信息。输入框本体复用共享 .textarea,这里只给提示文案留出间距。 */
-.vrep-page .replace-voice-row {
-  display: flex;
-  align-items: center;
-  flex-wrap: wrap;
-  gap: 8px;
-  margin-top: 10px;
+/* 使用限制提示。复用共享 .tip,这里只把配色换成警示色(用现有 token,不新建色值)。 */
+.vrep-page .replace-limit-tip {
+  margin-top: 22px;
+  background: var(--honey-bg);
+  border-style: solid;
+  border-color: var(--honey-bd);
+  color: var(--black-alpha-72);
 }

-.vrep-page .replace-voice-label {
-  font-size: 13px;
-  font-weight: 500;
-  color: var(--accent-black);
-  margin-right: 2px;
+.vrep-page .replace-limit-tip strong {
+  color: var(--accent-honey);
 }

 .vrep-page .replace-brief-hint {