请教：在中文输入环境下，目前推荐哪款本地 ASR 模型？

[EthanYoQ]

各位好，

我想了解一下在当前的 OpenLess 本地 ASR 支持中，针对纯中文及中英混输场景，大家最推荐哪一个本地模型？

在实际使用时，我主要关注以下几点：

1. 识别准确率与标点符号推断（尤其是中英夹杂与日常口语）
2. 响应延迟 / 实时性（打字录入的等待体感）
3. 资源占用（CPU / GPU 显存负载）

希望听听开发者和社区同学在实际测试与日常使用中的选型建议（例如 SenseVoice-Small vs. Whisper 系列 vs. Paraformer）。非常感谢！

Hi all,

I would like to ask for recommendations on the best local ASR engine/model for Chinese and mixed Chinese-English voice dictation in OpenLess.

Key aspects I am particularly interested in:

1. Recognition Accuracy & Punctuation (especially for casual phrasing and code-switching).
2. Inference Latency (responsiveness during daily typing).
3. Hardware Efficiency (CPU/GPU resource consumption).

What has been your experience or benchmark results across options like SenseVoice-Small, Whisper variants (e.g., small/medium/large-v3), or Paraformer?

Any insights or recommended configurations would be greatly appreciated!