Open-Source Audio Model Searches Apollo 11 Recordings

BanghuaZ · x · 2026-07-12

Open_MOSS has released **MOSS-Transcribe-Diarize**, an open-source **0.9B** model under the **Apache 2.0** license. This model combines **transcription**, **diarization**, and **timestamping** into a single pass, with its core advantage being suitability for **long audio inputs**. The provided real-world test case involves: - Processing **174 hours** of **Apollo 11** mission audio into searchable content - A total cost of just **$9.46** - Using original NASA recordings preserved by the Internet Archive - Processing a total of **103** audio segments - Completion via a single Hugging Face Job in **3.8 hours**, at a speed of roughly **47x realtime** - Generating **45,355** timestamped speaker segments The author notes that users can now search these mission recordings by "who said what and when."

Related event: Open_MOSS Releases Apollo Audio Transcription Model(2 posts)→

Original post →

More from Multimodal

Multimodal channel →