@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…
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.
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…
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.
@XieZhifei14110: Stop using Whisper for ASR ! open sourcing Mega-ASR — the first full-scenario SOTA industrial-grade ASR model, built fo…
Open sourcing Mega-ASR, a full-scenario SOTA industrial-grade ASR model designed for challenging audio conditions like far-field and noise, outperforming existing open and closed models by 10-30% on real-world benchmarks.
@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.
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...
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
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.