@QingQ77: Decouple Alibaba DAMO Academy's ZipEnhancer noise reduction model from the ModelScope pipeline and package it as a high-performance FastAPI denoising service. https://github.com/gyj1201/zipEnhancer… Alibaba DAMO Academy's Z…

X AI KOLs Timeline Tools

Summary

This project decouples Alibaba DAMO Academy's ZipEnhancer noise reduction model from the ModelScope pipeline, rewrites the inference logic in pure PyTorch, and packages it as a FastAPI service. It supports FP16 half-precision and long audio segmentation, providing multiple noise reduction model switching and API interfaces.

Decouple Alibaba DAMO Academy's ZipEnhancer noise reduction model from the ModelScope pipeline and package it as a high-performance FastAPI denoising service. https://github.com/gyj1201/zipEnhancer… Alibaba DAMO Academy's ZipEnhancer noise reduction model is typically only usable within ModelScope's pipeline, which is opaque and slow. This project extracts the model weights, rewrites the inference logic in pure PyTorch, and packages it as a FastAPI service. With FP16 half-precision, memory usage is reduced by about 40%; using a segmentation strategy with a 4-second sliding window and 75% overlap, it can handle audio of any length without running out of memory.
Original Article
View Cached Full Text

Cached at: 06/09/26, 10:45 AM

______ _____ _ |__ () __ | | __ | | __ _ _ __ ___ ___ _ __ / /| | ’ | | | ’ | ’_ \ / ` | ’ \ / / _ \ ’| / /| | |) | || | | | | | | (| | | | | (| __/ | /|| ./|___|| ||| ||_,|| ||__|| ||

Similar Articles

@FeitengLi: Actually, these problems can be well solved: 1. Ditch whisper, switch to an ASR model. Qwen3-ASR is great with few hallucinations, and there are other ASR options. Whisper has many hallucinations and requires 30s segments. Qwen3-ASR gets more accurate with longer audio, supporting up to 20…

X AI KOLs Timeline

Recommends using Qwen3-ASR instead of Whisper to reduce hallucinations, using LattifAI tools for precise audio-text alignment and subtitle generation, and introducing their own OmniVAD-Kit project for voice activity detection.

@nini_incrypto_: Headroom slashes LLM token costs by 95%! 1. True zero-code change: provides a proxy mode — any programming language can seamlessly integrate by just changing a port. 2. Full-throughput compression: automatically compresses tool outputs, runtime logs, RAG knowledge base chunks, and dense chat histories.

X AI KOLs Timeline

Headroom is a context compression layer that cuts AI agent token costs by 60–95%, supports a zero-code-change proxy mode, and does not degrade model response quality.

@MaxForAI: If you are working on voice agents, you should try this project. A team from NTU, NUS, and Shanghai AI Lab released: Mega-ASR. This fully open-source ASR is built on Qwen3-ASR, aiming to break the long-standing bottleneck of ASR performance in noisy, reverberant, or other impaired real-world environments...

X AI KOLs Timeline

NTU, NUS, and Shanghai AI Lab jointly released Mega-ASR, a fully open-source ASR model built on Qwen3-ASR. Using the Voices-in-the-Wild-2M dataset and progressive acoustic-to-semantic optimization, it achieves up to 30% relative Word Error Rate (WER) reduction in real-world noisy environments. With only 1.7B parameters, it enables efficient inference on consumer-grade hardware.

@leeoxiang: Really need a precise timestamp alignment service

X AI KOLs Following

Feiteng Li announces the release of EdgeSpeak, a local-first, privacy-preserving accurate transcription service. It supports drag-and-drop audio/video or microphone recording transcription, semantic segmentation, word-level timestamps, and export to JSON/SRT/MD. It is compatible with OpenAI Audio API, CLI, SKILL, and MCP workflows.