mirror of
https://github.com/linyqh/NarratoAI.git
synced 2026-07-22 14:08:19 +00:00
feat(subtitle): 新增 FireRedASR2 本地 ASR 后端支持
添加 FireRedASR2 本地 ASR 转写后端的完整支持： 1. 新增配置参数与数据模型字段 2. 更新示例配置文件，添加默认本地服务地址 3. 完善任务服务中的转写逻辑，支持 FireRedASR 后端 4. 更新 WebUI 界面，新增对应配置选项 5. 补充中英文多语言翻译 6. 新增本地 FireRedASR 服务的单元测试
This commit is contained in:
viccy
parent
e6e39d2dcd
commit
34d5532119
@ -201,6 +201,7 @@ class VideoClipParams(BaseModel):
|
||||
subtitle_auto_transcribe_enabled: bool = False
|
||||
subtitle_auto_transcribe_backend: str = "local"
|
||||
subtitle_auto_transcribe_api_url: str = ""
|
||||
subtitle_auto_transcribe_firered_api_url: str = ""
|
||||
subtitle_auto_transcribe_api_key: str = ""
|
||||
subtitle_auto_transcribe_hotword: str = ""
|
||||
subtitle_auto_transcribe_enable_spk: bool = False
|
||||
|
||||
@ -24,6 +24,7 @@ TRANSCRIPTION_URL = f"{DASHSCOPE_BASE_URL}/api/v1/services/audio/asr/transcripti
|
||||
TASK_URL_TEMPLATE = f"{DASHSCOPE_BASE_URL}/api/v1/tasks/{{task_id}}"
|
||||
MODEL_NAME = "fun-asr"
|
||||
LOCAL_FUN_ASR_API_URL = "http://127.0.0.1:7860"
|
||||
LOCAL_FIRERED_ASR_API_URL = "http://127.0.0.1:7867"
|
||||
TERMINAL_FAILED_STATUSES = {"FAILED", "CANCELED", "UNKNOWN"}
|
||||
PUNCTUATION_BREAKS = set("，。！？；,.!?;")
|
||||
|
||||
@ -131,7 +132,11 @@ def _absolute_local_download_url(api_url: str, download_url: str) -> str:
|
||||
return urljoin(f"{_local_base_url(api_url)}/", download_url)
|
||||
|
||||
|
||||
def _raise_for_local_http(response: requests.Response, action: str) -> None:
|
||||
def _raise_for_local_http(
|
||||
response: requests.Response,
|
||||
action: str,
|
||||
service_name: str = "本地 FunASR-Pack 服务",
|
||||
) -> None:
|
||||
status_code = getattr(response, "status_code", 200)
|
||||
if status_code and status_code >= 400:
|
||||
detail = ""
|
||||
@ -142,16 +147,20 @@ def _raise_for_local_http(response: requests.Response, action: str) -> None:
|
||||
except Exception:
|
||||
detail = ""
|
||||
suffix = f": {detail}" if detail else ""
|
||||
raise FunAsrError(f"{action}失败{suffix}，请确认本地 FunASR-Pack 服务可用")
|
||||
raise FunAsrError(f"{action}失败{suffix}，请确认{service_name}可用")
|
||||
|
||||
try:
|
||||
response.raise_for_status()
|
||||
except Exception as exc:
|
||||
raise FunAsrError(f"{action}失败，请确认本地 FunASR-Pack 服务可用") from exc
|
||||
raise FunAsrError(f"{action}失败，请确认{service_name}可用") from exc
|
||||
|
||||
|
||||
def _local_json(response: requests.Response, action: str) -> dict[str, Any]:
|
||||
_raise_for_local_http(response, action)
|
||||
def _local_json(
|
||||
response: requests.Response,
|
||||
action: str,
|
||||
service_name: str = "本地 FunASR-Pack 服务",
|
||||
) -> dict[str, Any]:
|
||||
_raise_for_local_http(response, action, service_name=service_name)
|
||||
try:
|
||||
data = response.json()
|
||||
except Exception as exc:
|
||||
@ -520,6 +529,19 @@ def request_local_fun_asr_health(api_url: str = LOCAL_FUN_ASR_API_URL, session=r
|
||||
return _local_json(response, "检查本地 FunASR-Pack 服务")
|
||||
|
||||
|
||||
def request_local_firered_asr_health(
|
||||
api_url: str = LOCAL_FIRERED_ASR_API_URL,
|
||||
session=requests,
|
||||
) -> dict[str, Any]:
|
||||
"""Fetch FireRedASR2-AED-Pack health metadata from the local service."""
|
||||
response = _session_get(session, f"{_local_base_url(api_url)}/health", timeout=10)
|
||||
return _local_json(
|
||||
response,
|
||||
"检查本地 FireRedASR2-AED-Pack 服务",
|
||||
service_name="本地 FireRedASR2-AED-Pack 服务",
|
||||
)
|
||||
|
||||
|
||||
def request_local_fun_asr(
|
||||
local_file: str,
|
||||
api_url: str = LOCAL_FUN_ASR_API_URL,
|
||||
@ -548,21 +570,61 @@ def request_local_fun_asr(
|
||||
return _local_json(response, "调用本地 FunASR-Pack ASR API")
|
||||
|
||||
|
||||
def request_local_firered_asr(
|
||||
local_file: str,
|
||||
api_url: str = LOCAL_FIRERED_ASR_API_URL,
|
||||
enable_vad: Optional[bool] = True,
|
||||
enable_lid: Optional[bool] = True,
|
||||
enable_punc: Optional[bool] = True,
|
||||
return_timestamp: Optional[bool] = True,
|
||||
timeout: float = 600.0,
|
||||
session=requests,
|
||||
) -> dict[str, Any]:
|
||||
"""Call the local FireRedASR2-AED-Pack `/asr` API and return its JSON result."""
|
||||
_require_local_file(local_file)
|
||||
data: dict[str, str] = {}
|
||||
options = {
|
||||
"enable_vad": enable_vad,
|
||||
"enable_lid": enable_lid,
|
||||
"enable_punc": enable_punc,
|
||||
"return_timestamp": return_timestamp,
|
||||
}
|
||||
for key, value in options.items():
|
||||
if value is not None:
|
||||
data[key] = "true" if value else "false"
|
||||
|
||||
with open(local_file, "rb") as file_obj:
|
||||
files = {"file": (_safe_upload_name(local_file), file_obj)}
|
||||
response = _session_post(
|
||||
session,
|
||||
_local_asr_url(api_url),
|
||||
data=data,
|
||||
files=files,
|
||||
timeout=timeout,
|
||||
)
|
||||
return _local_json(
|
||||
response,
|
||||
"调用本地 FireRedASR2-AED-Pack ASR API",
|
||||
service_name="本地 FireRedASR2-AED-Pack 服务",
|
||||
)
|
||||
|
||||
|
||||
def download_local_srt(
|
||||
download_url: str,
|
||||
api_url: str = LOCAL_FUN_ASR_API_URL,
|
||||
subtitle_file: str = "",
|
||||
session=requests,
|
||||
service_name: str = "本地 FunASR-Pack 服务",
|
||||
) -> str:
|
||||
"""Download an SRT exposed by FunASR-Pack and save it as a NarratoAI subtitle."""
|
||||
absolute_url = _absolute_local_download_url(api_url, download_url)
|
||||
if not absolute_url:
|
||||
raise FunAsrError("本地 FunASR-Pack 结果缺少 SRT 下载地址")
|
||||
response = _session_get(session, absolute_url, timeout=60)
|
||||
_raise_for_local_http(response, "下载本地 FunASR-Pack SRT")
|
||||
_raise_for_local_http(response, "下载本地 SRT", service_name=service_name)
|
||||
srt_content = _response_text(response)
|
||||
if not srt_content.strip():
|
||||
raise FunAsrError("本地 FunASR-Pack 返回了空 SRT")
|
||||
raise FunAsrError(f"{service_name}返回了空 SRT")
|
||||
return write_srt_file(srt_content, subtitle_file)
|
||||
|
||||
|
||||
@ -665,6 +727,45 @@ def local_fun_asr_result_to_srt(
|
||||
return "\n".join(lines).rstrip() + "\n"
|
||||
|
||||
|
||||
def firered_asr_result_to_srt(result_json: dict[str, Any]) -> str:
|
||||
"""Convert a FireRedASR2-AED-Pack JSON response into SRT when no SRT URL is returned."""
|
||||
blocks: list[dict[str, Any]] = []
|
||||
sentences = result_json.get("sentences")
|
||||
if isinstance(sentences, list):
|
||||
for sentence in sentences:
|
||||
if not isinstance(sentence, dict):
|
||||
continue
|
||||
text = str(sentence.get("text") or "").strip()
|
||||
if not text:
|
||||
continue
|
||||
start = sentence.get("start_ms", sentence.get("begin_time", sentence.get("start_time", 0)))
|
||||
end = sentence.get("end_ms", sentence.get("end_time"))
|
||||
start_ms = _timestamp_ms(start, "firered.sentence.start_ms")
|
||||
end_ms = _timestamp_ms(end, "firered.sentence.end_ms") if end is not None else start_ms + 500
|
||||
blocks.append({"start": start_ms, "end": end_ms, "text": text})
|
||||
|
||||
if not blocks:
|
||||
return local_fun_asr_result_to_srt(result_json)
|
||||
|
||||
lines = []
|
||||
for index, block in enumerate(blocks, start=1):
|
||||
lines.append(_srt_block(index, block["start"], block["end"], block["text"]))
|
||||
return "\n".join(lines).rstrip() + "\n"
|
||||
|
||||
|
||||
def _get_local_srt_download_url(result_json: dict[str, Any]) -> str:
|
||||
downloads = result_json.get("downloads") or {}
|
||||
if isinstance(downloads, dict):
|
||||
download_url = downloads.get("srt")
|
||||
if download_url:
|
||||
return str(download_url)
|
||||
for key in ("srt_url", "srt_download_url", "download_url"):
|
||||
download_url = result_json.get(key)
|
||||
if download_url:
|
||||
return str(download_url)
|
||||
return ""
|
||||
|
||||
|
||||
def create_with_local_fun_asr(
|
||||
local_file: str,
|
||||
subtitle_file: str = "",
|
||||
@ -689,8 +790,7 @@ def create_with_local_fun_asr(
|
||||
if isinstance(srt_file, str) and srt_file and os.path.isfile(srt_file):
|
||||
output_file = copy_srt_file(srt_file, subtitle_file)
|
||||
else:
|
||||
downloads = result_json.get("downloads") or {}
|
||||
download_url = downloads.get("srt") if isinstance(downloads, dict) else ""
|
||||
download_url = _get_local_srt_download_url(result_json)
|
||||
if download_url:
|
||||
output_file = download_local_srt(
|
||||
download_url,
|
||||
@ -710,6 +810,56 @@ def create_with_local_fun_asr(
|
||||
raise FunAsrError("本地 FunASR-Pack 字幕转写失败，请检查服务地址、文件或模型状态") from exc
|
||||
|
||||
|
||||
def create_with_local_firered_asr(
|
||||
local_file: str,
|
||||
subtitle_file: str = "",
|
||||
api_url: str = LOCAL_FIRERED_ASR_API_URL,
|
||||
enable_vad: Optional[bool] = True,
|
||||
enable_lid: Optional[bool] = True,
|
||||
enable_punc: Optional[bool] = True,
|
||||
return_timestamp: Optional[bool] = True,
|
||||
timeout: float = 600.0,
|
||||
session=requests,
|
||||
) -> Optional[str]:
|
||||
"""Create an SRT file through a locally running FireRedASR2-AED-Pack API."""
|
||||
service_name = "本地 FireRedASR2-AED-Pack 服务"
|
||||
try:
|
||||
result_json = request_local_firered_asr(
|
||||
local_file=local_file,
|
||||
api_url=api_url,
|
||||
enable_vad=enable_vad,
|
||||
enable_lid=enable_lid,
|
||||
enable_punc=enable_punc,
|
||||
return_timestamp=return_timestamp,
|
||||
timeout=timeout,
|
||||
session=session,
|
||||
)
|
||||
|
||||
srt_file = result_json.get("srt_file")
|
||||
if isinstance(srt_file, str) and srt_file and os.path.isfile(srt_file):
|
||||
output_file = copy_srt_file(srt_file, subtitle_file)
|
||||
else:
|
||||
download_url = _get_local_srt_download_url(result_json)
|
||||
if download_url:
|
||||
output_file = download_local_srt(
|
||||
download_url,
|
||||
api_url=api_url,
|
||||
subtitle_file=subtitle_file,
|
||||
session=session,
|
||||
service_name=service_name,
|
||||
)
|
||||
else:
|
||||
srt_content = firered_asr_result_to_srt(result_json)
|
||||
output_file = write_srt_file(srt_content, subtitle_file)
|
||||
|
||||
logger.info(f"本地 FireRedASR2-AED-Pack 字幕文件已生成: {output_file}")
|
||||
return output_file
|
||||
except FunAsrError:
|
||||
raise
|
||||
except Exception as exc:
|
||||
raise FunAsrError("本地ASR字幕转写失败，请检查 FireRedASR2-AED-Pack 服务地址、文件或模型状态") from exc
|
||||
|
||||
|
||||
def create_with_fun_asr(
|
||||
local_file: str,
|
||||
subtitle_file: str = "",
|
||||
|
||||
@ -24,7 +24,7 @@ def _is_auto_transcription_enabled(params: VideoClipParams) -> bool:
|
||||
|
||||
def _get_auto_transcription_backend(params: VideoClipParams) -> str:
|
||||
backend = str(getattr(params, "subtitle_auto_transcribe_backend", "") or "").strip().lower()
|
||||
if backend not in {"local", "bailian"}:
|
||||
if backend not in {"local", "firered", "bailian"}:
|
||||
backend = "local"
|
||||
return backend
|
||||
|
||||
@ -80,6 +80,19 @@ def _transcribe_final_video(task_id: str, video_path: str, params: VideoClipPara
|
||||
hotword=str(getattr(params, "subtitle_auto_transcribe_hotword", "") or "").strip(),
|
||||
enable_spk=bool(getattr(params, "subtitle_auto_transcribe_enable_spk", False)),
|
||||
)
|
||||
elif backend == "firered":
|
||||
api_url = str(
|
||||
getattr(params, "subtitle_auto_transcribe_firered_api_url", "")
|
||||
or config.fun_asr.get("firered_api_url", fun_asr_subtitle.LOCAL_FIRERED_ASR_API_URL)
|
||||
).strip()
|
||||
if not api_url:
|
||||
raise ValueError("请先输入本地ASR API 地址")
|
||||
|
||||
generated_path = fun_asr_subtitle.create_with_local_firered_asr(
|
||||
local_file=video_path,
|
||||
subtitle_file=subtitle_file,
|
||||
api_url=api_url,
|
||||
)
|
||||
else:
|
||||
api_key = str(
|
||||
getattr(params, "subtitle_auto_transcribe_api_key", "")
|
||||
|
||||
@ -481,6 +481,91 @@ class LocalFunAsrServiceTests(unittest.TestCase):
|
||||
self.assertIn("世界。", srt)
|
||||
|
||||
|
||||
class LocalFireRedAsrServiceTests(unittest.TestCase):
|
||||
def test_request_local_firered_asr_posts_file_and_options(self):
|
||||
class LocalSession:
|
||||
def __init__(self):
|
||||
self.calls = []
|
||||
|
||||
def post(self, url, **kwargs):
|
||||
self.calls.append(("POST", url, kwargs))
|
||||
return FakeResponse({"text": "你好", "srt_url": "/outputs/out.srt"})
|
||||
|
||||
with tempfile.TemporaryDirectory() as tmp_dir:
|
||||
local_file = Path(tmp_dir) / "audio.wav"
|
||||
local_file.write_bytes(b"audio")
|
||||
session = LocalSession()
|
||||
|
||||
result = fasr.request_local_firered_asr(
|
||||
str(local_file),
|
||||
api_url="127.0.0.1:7867",
|
||||
enable_vad=True,
|
||||
enable_lid=False,
|
||||
enable_punc=True,
|
||||
return_timestamp=True,
|
||||
timeout=456,
|
||||
session=session,
|
||||
)
|
||||
|
||||
self.assertEqual("你好", result["text"])
|
||||
self.assertEqual("POST", session.calls[0][0])
|
||||
self.assertEqual("http://127.0.0.1:7867/asr", session.calls[0][1])
|
||||
self.assertEqual(
|
||||
{
|
||||
"enable_vad": "true",
|
||||
"enable_lid": "false",
|
||||
"enable_punc": "true",
|
||||
"return_timestamp": "true",
|
||||
},
|
||||
session.calls[0][2]["data"],
|
||||
)
|
||||
self.assertEqual(456, session.calls[0][2]["timeout"])
|
||||
self.assertIn("file", session.calls[0][2]["files"])
|
||||
|
||||
def test_create_with_local_firered_asr_downloads_srt_url(self):
|
||||
class LocalSession:
|
||||
def __init__(self):
|
||||
self.calls = []
|
||||
|
||||
def post(self, url, **kwargs):
|
||||
self.calls.append(("POST", url, kwargs))
|
||||
return FakeResponse({"text": "你好", "srt_url": "/outputs/result.srt"})
|
||||
|
||||
def get(self, url, **kwargs):
|
||||
self.calls.append(("GET", url, kwargs))
|
||||
return FakeResponse(text="1\n00:00:00,000 --> 00:00:01,000\n你好\n")
|
||||
|
||||
with tempfile.TemporaryDirectory() as tmp_dir:
|
||||
local_file = Path(tmp_dir) / "audio.wav"
|
||||
local_file.write_bytes(b"audio")
|
||||
subtitle_file = Path(tmp_dir) / "out.srt"
|
||||
session = LocalSession()
|
||||
|
||||
result_path = fasr.create_with_local_firered_asr(
|
||||
str(local_file),
|
||||
subtitle_file=str(subtitle_file),
|
||||
api_url="http://127.0.0.1:7867",
|
||||
session=session,
|
||||
)
|
||||
|
||||
self.assertEqual(str(subtitle_file), result_path)
|
||||
self.assertEqual("http://127.0.0.1:7867/outputs/result.srt", session.calls[1][1])
|
||||
self.assertIn("你好", subtitle_file.read_text(encoding="utf-8"))
|
||||
|
||||
def test_firered_asr_result_to_srt_uses_sentence_timestamps(self):
|
||||
result = {
|
||||
"sentences": [
|
||||
{"text": "你好。", "start_ms": 40, "end_ms": 900},
|
||||
{"text": "欢迎观看。", "start_ms": 900, "end_ms": 2100},
|
||||
]
|
||||
}
|
||||
|
||||
srt = fasr.firered_asr_result_to_srt(result)
|
||||
|
||||
self.assertIn("1\n00:00:00,040 --> 00:00:00,900\n你好。", srt)
|
||||
self.assertIn("2\n00:00:00,900 --> 00:00:02,100\n欢迎观看。", srt)
|
||||
|
||||
|
||||
class FunAsrConfigTests(unittest.TestCase):
|
||||
def test_save_config_persists_fun_asr_section(self):
|
||||
original_config_file = cfg.config_file
|
||||
@ -503,6 +588,7 @@ class FunAsrConfigTests(unittest.TestCase):
|
||||
config_data = tomllib.loads(Path("config.example.toml").read_text(encoding="utf-8"))
|
||||
self.assertEqual("local", config_data["fun_asr"]["backend"])
|
||||
self.assertEqual("http://127.0.0.1:7860", config_data["fun_asr"]["api_url"])
|
||||
self.assertEqual("http://127.0.0.1:7867", config_data["fun_asr"]["firered_api_url"])
|
||||
self.assertEqual("fun-asr", config_data["fun_asr"]["model"])
|
||||
self.assertIn("api_key", config_data["fun_asr"])
|
||||
|
||||
|
||||
@ -110,10 +110,11 @@
|
||||
|
||||
[fun_asr]
|
||||
# Fun-ASR 字幕转录配置
|
||||
# backend = "local" 使用本地 FunASR-Pack API；backend = "bailian" 使用阿里百炼在线 fun-asr
|
||||
# backend = "local" 使用本地 FunASR-Pack API；backend = "firered" 使用本地 FireRedASR2-AED-Pack API；backend = "bailian" 使用阿里百炼在线 fun-asr
|
||||
auto_transcribe_enabled = false
|
||||
backend = "local"
|
||||
api_url = "http://127.0.0.1:7860"
|
||||
firered_api_url = "http://127.0.0.1:7867"
|
||||
hotword = ""
|
||||
enable_spk = false
|
||||
# 使用阿里百炼在线 fun-asr 时，访问 https://bailian.console.aliyun.com/?tab=model#/api-key 获取 API Key
|
||||
|
||||
@ -996,11 +996,12 @@ def render_fun_asr_transcription(tr):
|
||||
|
||||
backend_options = {
|
||||
tr("Local FunASR-Pack API"): "local",
|
||||
tr("Local FireRedASR API"): "firered",
|
||||
tr("Ali Bailian Online Fun-ASR"): "bailian",
|
||||
tr("上传字幕文件"): "upload",
|
||||
}
|
||||
saved_backend = str(config.fun_asr.get("backend", "")).strip().lower()
|
||||
if saved_backend not in {"local", "bailian", "upload"}:
|
||||
if saved_backend not in {"local", "firered", "bailian", "upload"}:
|
||||
saved_backend = (
|
||||
"bailian"
|
||||
if config.fun_asr.get("api_key") and not config.fun_asr.get("api_url")
|
||||
@ -1012,6 +1013,7 @@ def render_fun_asr_transcription(tr):
|
||||
backend = saved_backend
|
||||
api_key = ""
|
||||
api_url = config.fun_asr.get("api_url", fun_asr_subtitle.LOCAL_FUN_ASR_API_URL)
|
||||
firered_api_url = config.fun_asr.get("firered_api_url", fun_asr_subtitle.LOCAL_FIRERED_ASR_API_URL)
|
||||
hotword = config.fun_asr.get("hotword", "")
|
||||
enable_spk = bool(config.fun_asr.get("enable_spk", False))
|
||||
media_paths = _selected_video_paths()
|
||||
@ -1020,11 +1022,10 @@ def render_fun_asr_transcription(tr):
|
||||
|
||||
with subtitle_cols[0]:
|
||||
with st.expander(tr("Ali Bailian Fun-ASR Subtitle Transcription"), expanded=False):
|
||||
backend_label = st.radio(
|
||||
backend_label = st.selectbox(
|
||||
tr("Subtitle Processing Method"),
|
||||
options=backend_labels,
|
||||
index=backend_values.index(saved_backend),
|
||||
horizontal=True,
|
||||
key="fun_asr_backend",
|
||||
)
|
||||
backend = backend_options[backend_label]
|
||||
@ -1051,6 +1052,14 @@ def render_fun_asr_transcription(tr):
|
||||
help=tr("Enable speaker diarization Help"),
|
||||
key="fun_asr_enable_spk",
|
||||
)
|
||||
elif backend == "firered":
|
||||
st.caption(tr("Local FireRed-ASR upload caption"))
|
||||
firered_api_url = st.text_input(
|
||||
tr("Local FireRedASR API URL"),
|
||||
value=firered_api_url,
|
||||
help=tr("Local FireRedASR API URL Help"),
|
||||
key="fun_asr_firered_api_url",
|
||||
)
|
||||
else:
|
||||
st.caption(tr("Fun-ASR upload caption"))
|
||||
st.markdown(
|
||||
@ -1166,6 +1175,10 @@ def render_fun_asr_transcription(tr):
|
||||
clear_fun_asr_subtitle_state()
|
||||
st.error(tr("Please enter local FunASR-Pack API URL"))
|
||||
return
|
||||
if backend == "firered" and not str(firered_api_url).strip():
|
||||
clear_fun_asr_subtitle_state()
|
||||
st.error(tr("Please enter local FireRedASR API URL"))
|
||||
return
|
||||
missing_paths = [path for path in media_paths if not os.path.exists(path)]
|
||||
if not media_paths or missing_paths:
|
||||
clear_fun_asr_subtitle_state()
|
||||
@ -1184,22 +1197,25 @@ def render_fun_asr_transcription(tr):
|
||||
|
||||
config.fun_asr["backend"] = backend
|
||||
config.fun_asr["api_url"] = str(api_url).strip()
|
||||
config.fun_asr["firered_api_url"] = str(firered_api_url).strip()
|
||||
config.fun_asr["api_key"] = api_key.strip()
|
||||
config.fun_asr["hotword"] = str(hotword).strip()
|
||||
config.fun_asr["enable_spk"] = bool(enable_spk)
|
||||
config.fun_asr["model"] = "fun-asr"
|
||||
config.save_config()
|
||||
|
||||
spinner_text = (
|
||||
tr("Transcribing with local FunASR-Pack...")
|
||||
if backend == "local"
|
||||
else tr("Transcribing with Fun-ASR...")
|
||||
)
|
||||
if backend == "local":
|
||||
spinner_text = tr("Transcribing with local FunASR-Pack...")
|
||||
elif backend == "firered":
|
||||
spinner_text = tr("Transcribing with local FireRedASR...")
|
||||
else:
|
||||
spinner_text = tr("Transcribing with Fun-ASR...")
|
||||
with st.spinner(spinner_text):
|
||||
progress_bar = st.progress(0) if len(media_paths) > 1 else None
|
||||
generated_paths = []
|
||||
for index, media_path in enumerate(media_paths, start=1):
|
||||
subtitle_name = f"{os.path.splitext(os.path.basename(media_path))[0]}_fun_asr.srt"
|
||||
subtitle_suffix = "firered_asr" if backend == "firered" else "fun_asr"
|
||||
subtitle_name = f"{os.path.splitext(os.path.basename(media_path))[0]}_{subtitle_suffix}.srt"
|
||||
subtitle_path = _unique_file_path(utils.subtitle_dir(), subtitle_name)
|
||||
|
||||
if backend == "local":
|
||||
@ -1210,6 +1226,12 @@ def render_fun_asr_transcription(tr):
|
||||
hotword=str(hotword).strip(),
|
||||
enable_spk=bool(enable_spk),
|
||||
)
|
||||
elif backend == "firered":
|
||||
generated_path = fun_asr_subtitle.create_with_local_firered_asr(
|
||||
local_file=media_path,
|
||||
subtitle_file=subtitle_path,
|
||||
api_url=str(firered_api_url).strip(),
|
||||
)
|
||||
else:
|
||||
generated_path = fun_asr_subtitle.create_with_fun_asr(
|
||||
local_file=media_path,
|
||||
|
||||
@ -457,7 +457,7 @@ def render_subtitle_mask_settings(tr):
|
||||
|
||||
def _get_saved_auto_transcribe_backend():
|
||||
saved_backend = str(config.fun_asr.get("backend", "")).strip().lower()
|
||||
if saved_backend not in {"local", "bailian"}:
|
||||
if saved_backend not in {"local", "firered", "bailian"}:
|
||||
saved_backend = (
|
||||
"bailian"
|
||||
if config.fun_asr.get("api_key") and not config.fun_asr.get("api_url")
|
||||
@ -481,6 +481,7 @@ def render_auto_transcription_settings(tr):
|
||||
|
||||
backend = _get_saved_auto_transcribe_backend()
|
||||
api_url = config.fun_asr.get("api_url", fun_asr_subtitle.LOCAL_FUN_ASR_API_URL)
|
||||
firered_api_url = config.fun_asr.get("firered_api_url", fun_asr_subtitle.LOCAL_FIRERED_ASR_API_URL)
|
||||
hotword = config.fun_asr.get("hotword", "")
|
||||
enable_spk = bool(config.fun_asr.get("enable_spk", False))
|
||||
api_key = config.fun_asr.get("api_key", "")
|
||||
@ -488,6 +489,7 @@ def render_auto_transcription_settings(tr):
|
||||
if not auto_transcribe_enabled:
|
||||
st.session_state['subtitle_auto_transcribe_backend'] = backend
|
||||
st.session_state['subtitle_auto_transcribe_api_url'] = api_url
|
||||
st.session_state['subtitle_auto_transcribe_firered_api_url'] = firered_api_url
|
||||
st.session_state['subtitle_auto_transcribe_hotword'] = hotword
|
||||
st.session_state['subtitle_auto_transcribe_enable_spk'] = enable_spk
|
||||
st.session_state['subtitle_auto_transcribe_api_key'] = api_key
|
||||
@ -495,17 +497,17 @@ def render_auto_transcription_settings(tr):
|
||||
|
||||
backend_options = {
|
||||
tr("Local FunASR-Pack API"): "local",
|
||||
tr("Local FireRedASR API"): "firered",
|
||||
tr("Ali Bailian Online Fun-ASR"): "bailian",
|
||||
}
|
||||
backend_values = list(backend_options.values())
|
||||
backend_labels = list(backend_options.keys())
|
||||
|
||||
backend_label = st.radio(
|
||||
backend_label = st.selectbox(
|
||||
tr("Subtitle Processing Method"),
|
||||
options=backend_labels,
|
||||
index=backend_values.index(backend),
|
||||
horizontal=True,
|
||||
key="subtitle_auto_transcribe_backend_radio",
|
||||
key="subtitle_auto_transcribe_backend_select",
|
||||
)
|
||||
backend = backend_options[backend_label]
|
||||
|
||||
@ -529,6 +531,14 @@ def render_auto_transcription_settings(tr):
|
||||
help=tr("Enable speaker diarization Help"),
|
||||
key="subtitle_auto_transcribe_enable_spk_checkbox",
|
||||
)
|
||||
elif backend == "firered":
|
||||
st.caption(tr("Auto Transcription FireRed Caption"))
|
||||
firered_api_url = st.text_input(
|
||||
tr("Local FireRedASR API URL"),
|
||||
value=firered_api_url,
|
||||
help=tr("Local FireRedASR API URL Help"),
|
||||
key="subtitle_auto_transcribe_firered_api_url_input",
|
||||
)
|
||||
else:
|
||||
st.caption(tr("Auto Transcription Online Caption"))
|
||||
st.markdown(
|
||||
@ -546,6 +556,7 @@ def render_auto_transcription_settings(tr):
|
||||
|
||||
config.fun_asr["backend"] = backend
|
||||
config.fun_asr["api_url"] = str(api_url).strip()
|
||||
config.fun_asr["firered_api_url"] = str(firered_api_url).strip()
|
||||
config.fun_asr["api_key"] = str(api_key).strip()
|
||||
config.fun_asr["hotword"] = str(hotword).strip()
|
||||
config.fun_asr["enable_spk"] = bool(enable_spk)
|
||||
@ -553,6 +564,7 @@ def render_auto_transcription_settings(tr):
|
||||
|
||||
st.session_state['subtitle_auto_transcribe_backend'] = backend
|
||||
st.session_state['subtitle_auto_transcribe_api_url'] = str(api_url).strip()
|
||||
st.session_state['subtitle_auto_transcribe_firered_api_url'] = str(firered_api_url).strip()
|
||||
st.session_state['subtitle_auto_transcribe_api_key'] = str(api_key).strip()
|
||||
st.session_state['subtitle_auto_transcribe_hotword'] = str(hotword).strip()
|
||||
st.session_state['subtitle_auto_transcribe_enable_spk'] = bool(enable_spk)
|
||||
@ -692,6 +704,10 @@ def get_subtitle_params():
|
||||
'subtitle_auto_transcribe_api_url',
|
||||
config.fun_asr.get("api_url", "")
|
||||
),
|
||||
'subtitle_auto_transcribe_firered_api_url': st.session_state.get(
|
||||
'subtitle_auto_transcribe_firered_api_url',
|
||||
config.fun_asr.get("firered_api_url", "")
|
||||
),
|
||||
'subtitle_auto_transcribe_api_key': st.session_state.get(
|
||||
'subtitle_auto_transcribe_api_key',
|
||||
config.fun_asr.get("api_key", "")
|
||||
|
||||
@ -414,14 +414,19 @@
|
||||
"Ali Bailian Fun-ASR Subtitle Transcription": "Subtitle Processing",
|
||||
"Subtitle Processing Method": "Subtitle Processing Method",
|
||||
"Fun-ASR Backend": "Fun-ASR Backend",
|
||||
"Local FunASR-Pack API": "Local",
|
||||
"Ali Bailian Online Fun-ASR": "Online",
|
||||
"Local FunASR-Pack API": "FunASR (Local)",
|
||||
"Local FireRedASR API": "FireRedASR2 (Local)",
|
||||
"Ali Bailian Online Fun-ASR": "FunASR (Online)",
|
||||
"Local Fun-ASR upload caption": "The current video above will be converted to SRT subtitles through the locally running FunASR-Pack API.",
|
||||
"Local FireRed-ASR upload caption": "The current video above will be converted to SRT subtitles through the locally running FireRedASR2-AED-Pack API.",
|
||||
"Fun-ASR upload caption": "The current video above will be uploaded to temporary Ali Bailian storage and converted to SRT subtitles with fun-asr.",
|
||||
"Auto Transcription Local Caption": "After the final video is merged, it will be converted to SRT subtitles through the locally running FunASR-Pack API.",
|
||||
"Auto Transcription FireRed Caption": "After the final video is merged, it will be converted to SRT subtitles through the locally running FireRedASR2-AED-Pack API.",
|
||||
"Auto Transcription Online Caption": "After the final video is merged, it will be uploaded to temporary Ali Bailian storage and converted to SRT subtitles with fun-asr.",
|
||||
"Local FunASR-Pack API URL": "Local FunASR-Pack API URL",
|
||||
"Local FunASR-Pack API URL Help": "For example, http://127.0.0.1:7860. A full /asr endpoint URL is also supported.",
|
||||
"Local FireRedASR API URL": "Local ASR API URL",
|
||||
"Local FireRedASR API URL Help": "For example, http://127.0.0.1:7867. A full /asr endpoint URL is also supported.",
|
||||
"Fun-ASR Hotword": "Hotword",
|
||||
"Fun-ASR Hotword Help": "Optional hotwords passed to the local FunASR-Pack API.",
|
||||
"Enable speaker diarization": "Enable speaker diarization",
|
||||
@ -439,8 +444,10 @@
|
||||
"Calibrate subtitles": "Calibrate Subtitles",
|
||||
"Please enter Ali Bailian API Key": "Please enter the Ali Bailian API Key first",
|
||||
"Please enter local FunASR-Pack API URL": "Please enter the local FunASR-Pack API URL first",
|
||||
"Please enter local FireRedASR API URL": "Please enter the local ASR API URL first",
|
||||
"Please upload media to transcribe": "Please upload the audio or video file to transcribe first",
|
||||
"Transcribing with local FunASR-Pack...": "Transcribing subtitles with local FunASR-Pack, please wait...",
|
||||
"Transcribing with local FireRedASR...": "Transcribing subtitles with local ASR, please wait...",
|
||||
"Transcribing with Fun-ASR...": "Transcribing subtitles with Ali Bailian Fun-ASR, please wait...",
|
||||
"Fun-ASR failed without subtitle file": "Fun-ASR transcription failed: no subtitle file was generated",
|
||||
"Subtitle transcription succeeded": "Subtitle transcription succeeded: {file}",
|
||||
|
||||
@ -365,14 +365,19 @@
|
||||
"Ali Bailian Fun-ASR Subtitle Transcription": "字幕处理",
|
||||
"Subtitle Processing Method": "字幕处理方式",
|
||||
"Fun-ASR Backend": "Fun-ASR 后端",
|
||||
"Local FunASR-Pack API": "本地转写",
|
||||
"Ali Bailian Online Fun-ASR": "在线转写",
|
||||
"Local FunASR-Pack API": "FunASR(本地部署)",
|
||||
"Local FireRedASR API": "FireRedASR2(本地部署)",
|
||||
"Ali Bailian Online Fun-ASR": "FunASR(在线服务)",
|
||||
"Local Fun-ASR upload caption": "将使用上方当前视频，通过本机运行的 FunASR-Pack API 生成 SRT 字幕。",
|
||||
"Local FireRed-ASR upload caption": "将使用上方当前视频，通过本机运行的 FireRedASR2-AED-Pack API 生成 SRT 字幕。",
|
||||
"Fun-ASR upload caption": "将使用上方当前视频，自动上传到阿里百炼临时存储并通过 fun-asr 生成 SRT 字幕。",
|
||||
"Auto Transcription Local Caption": "将在最终视频合并完成后，通过本机运行的 FunASR-Pack API 生成 SRT 字幕。",
|
||||
"Auto Transcription FireRed Caption": "将在最终视频合并完成后，通过本机运行的 FireRedASR2-AED-Pack API 生成 SRT 字幕。",
|
||||
"Auto Transcription Online Caption": "将在最终视频合并完成后，自动上传到阿里百炼临时存储并通过 fun-asr 生成 SRT 字幕。",
|
||||
"Local FunASR-Pack API URL": "本地 FunASR-Pack API 地址",
|
||||
"Local FunASR-Pack API URL Help": "例如 http://127.0.0.1:7860；也可以直接填到 /asr 的完整地址。",
|
||||
"Local FireRedASR API URL": "本地ASR API 地址",
|
||||
"Local FireRedASR API URL Help": "例如 http://127.0.0.1:7867；也可以直接填到 /asr 的完整地址。",
|
||||
"Fun-ASR Hotword": "热词",
|
||||
"Fun-ASR Hotword Help": "可选，传给本地 FunASR-Pack 的热词参数。",
|
||||
"Enable speaker diarization": "启用说话人分段",
|
||||
@ -390,8 +395,10 @@
|
||||
"Calibrate subtitles": "校准字幕",
|
||||
"Please enter Ali Bailian API Key": "请先输入阿里百炼 API Key",
|
||||
"Please enter local FunASR-Pack API URL": "请先输入本地 FunASR-Pack API 地址",
|
||||
"Please enter local FireRedASR API URL": "请先输入本地ASR API 地址",
|
||||
"Please upload media to transcribe": "请先上传需要转录的音频或视频文件",
|
||||
"Transcribing with local FunASR-Pack...": "正在使用本地 FunASR-Pack 转写字幕，请稍候...",
|
||||
"Transcribing with local FireRedASR...": "正在使用本地ASR转写字幕，请稍候...",
|
||||
"Transcribing with Fun-ASR...": "正在使用阿里百炼 Fun-ASR 转写字幕，请稍候...",
|
||||
"Fun-ASR failed without subtitle file": "Fun-ASR 转写失败：未生成字幕文件",
|
||||
"Subtitle transcription succeeded": "字幕转写成功: {file}",
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user