Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

chore(release): prepare FunASR 1.4.21

zhifu gaoda07e5be75f4
diff --git a/README.md b/README.md
index 5aa159650..277870f0a 100644
--- a/README.md
+++ b/README.md
@@ -364,7 +364,8 @@ recordings with the same evaluation scope.
## What's new
- **MOSS-Transcribe-Diarize** brings long-form ASR, timestamps, and anonymous speaker labels to FunASR services, Docker, Kubernetes, vLLM/SGLang workflows, and FunClip. [Deploy MOSS ->](./docs/moss_transcribe_diarize.md)
-- **FunASR 1.4.20** preserves the original optional speaker dependency failure in a chained `ImportError` when `spk_model` is requested, while ASR-only construction remains available. [Fix #3819](https://github.com/modelscope/FunASR/pull/3819). Install with `python -m pip install -U "funasr==1.4.20"`. [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
+- **FunASR 1.4.21** preserves the streaming punctuation cache on empty or whitespace chunks, returning an empty result without touching cached words or the embedding path; nonempty inference is unchanged. [Fix #3824](https://github.com/modelscope/FunASR/pull/3824). Install with `python -m pip install -U "funasr==1.4.21"`. [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.21)
+ **FunASR 1.4.20** preserves the original optional speaker dependency failure in a chained `ImportError` when `spk_model` is requested, while ASR-only construction remains available. [Fix #3819](https://github.com/modelscope/FunASR/pull/3819). [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
**FunASR 1.4.19** handles empty-token Paraformer batches, enforces `max_speakers` for automatic clustering, and adds validated JSON-object `--hf_overrides` support to realtime WebSocket serving. [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.19)
Previous release **FunASR 1.4.18** fixes cross-script hotword misreplacement, filtered-token timestamp alignment, and Chinese VAD sentence/speaker drift. [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.18)
**Earlier release: FunASR 1.4.17** adds explicit PCM16 input and improves timestamp, hotword, speaker-sample-rate, and long-segment streaming handling. [Release and verification scope ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.17)
diff --git a/README_ja.md b/README_ja.md
index 18445b4ba..13450d864 100644
--- a/README_ja.md
+++ b/README_ja.md
@@ -112,7 +112,8 @@ CER/WER をそろえて比較してください。オフラインのスループ
## 最新情報
- **MOSS-Transcribe-Diarize** を FunASR service、Docker、Kubernetes、vLLM/SGLang workflow、FunClip に統合し、長時間 ASR、timestamp、匿名 speaker label を一度に処理できます。[MOSS をデプロイ ->](./docs/moss_transcribe_diarize.md)
-- **FunASR 1.4.20** は `spk_model` 指定時に任意の話者依存モジュールを読み込めない場合、元の例外を連鎖した `ImportError` で報告します。ASR のみの構築は引き続き利用できます。[修正 #3819](https://github.com/modelscope/FunASR/pull/3819)。`python -m pip install -U "funasr==1.4.20"` でインストールできます。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
+- **FunASR 1.4.21** はストリーミング句読点処理の空または空白入力で cache を保持し、キャッシュ済み単語や embedding 経路に触れず空結果を返します。非空推論は変更されません。[修正 #3824](https://github.com/modelscope/FunASR/pull/3824)。`python -m pip install -U "funasr==1.4.21"` でインストールできます。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.21)
+ **FunASR 1.4.20** は `spk_model` 指定時に任意の話者依存モジュールを読み込めない場合、元の例外を連鎖した `ImportError` で報告します。ASR のみの構築は引き続き利用できます。[修正 #3819](https://github.com/modelscope/FunASR/pull/3819)。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
**FunASR 1.4.19** は Paraformer の空 token batch を処理し、自動クラスタリングの `max_speakers` 上限を適用します。リアルタイム WebSocket サービスには JSON object を検証する `--hf_overrides` を追加しました。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.19)
前のリリース **FunASR 1.4.18** は中国語とラテン文字が混在する hotword の誤置換、特殊 token 除去後の timestamp 対応、中国語 VAD 分割での文と話者のずれを修正しました。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.18)
以前のリリース **FunASR 1.4.17** は明示的な PCM16 入力を追加し、timestamp、hotword、speaker sample rate、長い segment の streaming 処理を改善しました。[リリースと検証範囲 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.17)
diff --git a/README_ko.md b/README_ko.md
index c05a06649..a4a8ece06 100644
--- a/README_ko.md
+++ b/README_ko.md
@@ -111,7 +111,8 @@ checkpoint/revision, 오디오 집합, 하드웨어, 배치, 워밍업, 측정
## 최신 소식
- **MOSS-Transcribe-Diarize**를 FunASR service, Docker, Kubernetes, vLLM/SGLang workflow, FunClip에 통합해 긴 오디오 ASR, timestamp, 익명 speaker label을 한 번에 처리합니다. [MOSS 배포 ->](./docs/moss_transcribe_diarize.md)
-- **FunASR 1.4.20**은 `spk_model` 사용 시 선택적 화자 의존성 가져오기가 실패하면 원래 예외를 연결한 `ImportError`를 제공합니다. ASR 전용 구성은 계속 사용할 수 있습니다. [수정 #3819](https://github.com/modelscope/FunASR/pull/3819). `python -m pip install -U "funasr==1.4.20"`로 설치할 수 있습니다. [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
+- **FunASR 1.4.21**은 스트리밍 문장부호 처리의 빈 입력이나 공백 입력에서 cache를 유지하고, 캐시된 단어나 embedding 경로를 건드리지 않은 채 빈 결과를 반환합니다. 비어 있지 않은 추론은 변경되지 않습니다. [수정 #3824](https://github.com/modelscope/FunASR/pull/3824). `python -m pip install -U "funasr==1.4.21"`로 설치할 수 있습니다. [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.21)
+ **FunASR 1.4.20**은 `spk_model` 사용 시 선택적 화자 의존성 가져오기가 실패하면 원래 예외를 연결한 `ImportError`를 제공합니다. ASR 전용 구성은 계속 사용할 수 있습니다. [수정 #3819](https://github.com/modelscope/FunASR/pull/3819). [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
**FunASR 1.4.19**는 Paraformer의 빈 token batch를 처리하고 자동 클러스터링에 `max_speakers` 상한을 적용합니다. 실시간 WebSocket 서비스에 JSON object를 검증하는 `--hf_overrides`도 추가했습니다. [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.19)
이전 릴리스 **FunASR 1.4.18**은 중국어와 라틴 문자가 섞인 hotword의 잘못된 치환, 특수 token 제거 후 timestamp 정렬, 중국어 VAD 분할의 문장 및 화자 불일치를 수정했습니다. [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.18)
그 이전 릴리스 **FunASR 1.4.17**은 명시적 PCM16 입력을 추가하고 timestamp, hotword, speaker sample rate, 긴 segment의 streaming 처리를 개선했습니다. [릴리스 및 검증 범위 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.17)
diff --git a/README_zh.md b/README_zh.md
index dc4b9f2f2..eccb4f714 100644
--- a/README_zh.md
+++ b/README_zh.md
@@ -157,7 +157,8 @@ checkpoint/revision、音频集、硬件、批量大小、预热、计时范围
## 最新动态
- **MOSS-Transcribe-Diarize** 已接入 FunASR 服务、Docker、Kubernetes、vLLM/SGLang 工作流和 FunClip,一次完成长音频转写、时间戳与匿名说话人标注。[部署 MOSS ->](./docs/moss_transcribe_diarize_zh.md)
-- **FunASR 1.4.20** 在指定 `spk_model` 而可选说话人依赖导入失败时,通过带原始异常链的 `ImportError` 提供诊断,同时保留仅 ASR 的构造路径。[修复 #3819](https://github.com/modelscope/FunASR/pull/3819)。安装:`python -m pip install -U "funasr==1.4.20"`。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
+- **FunASR 1.4.21** 在流式标点收到空白或空输入时保留 cache,并在不访问缓存词或 embedding 路径的情况下返回空结果;非空推理保持不变。[修复 #3824](https://github.com/modelscope/FunASR/pull/3824)。安装:`python -m pip install -U "funasr==1.4.21"`。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.21)
+ **FunASR 1.4.20** 在指定 `spk_model` 而可选说话人依赖导入失败时,通过带原始异常链的 `ImportError` 提供诊断,同时保留仅 ASR 的构造路径。[修复 #3819](https://github.com/modelscope/FunASR/pull/3819)。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)
**FunASR 1.4.19** 修复 Paraformer 空 token 批次处理,使自动聚类遵守 `max_speakers` 上限,并为实时 WebSocket 服务新增带 JSON 对象校验的 `--hf_overrides` 参数。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.19)
上一版本 **FunASR 1.4.18** 修复跨中英文热词误替换、过滤特殊符号后的时间戳对齐,以及中文 VAD 分段的句子与说话人错位。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.18)
**更早版本:FunASR 1.4.17** 新增显式 PCM16 输入,改进时间戳、热词、说话人采样率与长片段流式处理。[发布说明与验证范围 ->](https://github.com/modelscope/FunASR/releases/tag/v1.4.17)
diff --git a/docs/repository_roles.md b/docs/repository_roles.md
index e7f5b88d5..911a2b655 100644
--- a/docs/repository_roles.md
+++ b/docs/repository_roles.md
@@ -5,7 +5,7 @@ This document explains the responsibility boundaries, user entry points, and iss
> **Directional roadmap, not a release promise.**
> This document records shipped capabilities and active work, but does not commit to
> future version numbers or dates. The current Python release is
-> [`funasr==1.4.20`](https://github.com/modelscope/FunASR/releases/tag/v1.4.20).
+> [`funasr==1.4.21`](https://github.com/modelscope/FunASR/releases/tag/v1.4.21).
> Any future breaking release still requires a maintainer-approved milestone and
> migration plan.
diff --git a/docs/repository_roles_zh.md b/docs/repository_roles_zh.md
index 66721d7d9..6a847f170 100644
--- a/docs/repository_roles_zh.md
+++ b/docs/repository_roles_zh.md
@@ -4,7 +4,7 @@
> **方向性路线图,不是版本承诺。**
> 本文档记录已交付能力与正在推进的工作,但不承诺未来版本号或日期。当前 Python
-> 版本是 [`funasr==1.4.20`](https://github.com/modelscope/FunASR/releases/tag/v1.4.20)。
+> 版本是 [`funasr==1.4.21`](https://github.com/modelscope/FunASR/releases/tag/v1.4.21)。
> 任何未来的 breaking release 仍需 maintainer 确认 milestone 与迁移方案。
---
diff --git a/funasr/version.txt b/funasr/version.txt
index 771411b02..8cc8fe024 100644
--- a/funasr/version.txt
+++ b/funasr/version.txt
@@ -1 +1 @@
-1.4.20
+1.4.21
diff --git a/tests/test_release_version_contract.py b/tests/test_release_version_contract.py
index 1606e0a3d..57623d690 100644
--- a/tests/test_release_version_contract.py
+++ b/tests/test_release_version_contract.py
@@ -4,17 +4,16 @@ import pytest
ROOT = Path(__file__).resolve().parents[1]
-RELEASE_VERSION = "1.4.20"
+RELEASE_VERSION = "1.4.21"
README_FILES = ("README.md", "README_zh.md", "README_ja.md", "README_ko.md")
REPOSITORY_ROLE_FILES = ("docs/repository_roles.md", "docs/repository_roles_zh.md")
@pytest.mark.parametrize("name", README_FILES)
-def test_release_summary_preserves_optional_speaker_failure_scope(name):
+def test_release_summary_preserves_streaming_punctuation_cache(name):
summaries = [line for line in (ROOT / name).read_text(encoding="utf-8").splitlines()
if f"**FunASR {RELEASE_VERSION}**" in line]
assert len(summaries) == 1
- assert "spk_model" in summaries[0]
- assert "ImportError" in summaries[0]
- assert "https://github.com/modelscope/FunASR/pull/3819" in summaries[0]
+ assert "cache" in summaries[0].lower()
+ assert "https://github.com/modelscope/FunASR/pull/3824" in summaries[0]