RepoRadar
Leave a Tip
Item detail

Soul-AILab/SoulX-Transcriber

SoulX-Transcriber is an Apache-2.0, end-to-end multi-speaker speech transcription framework that jointly models *who* spoke, *when*, and *what* — speaker-attributed transcripts with timestamps, designed to handle overlapping speech and conversational dynamics that ASR systems usually miss. 250 stars, PyTorch, ships pretrained checkpoints and a Python SDK.

Open source →Back to radar
Score7.6
Popularity62.0
Risknone
TierSilver
Score breakdown
Usefulness7.0
Novelty9.0
Momentum6.0
Maturity6.8
Open-source/build8.4
Evidence7.2
Workflow potential8.0
Setup ease6.4

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Useful for researchers, podcasters, journalists, and product teams who need speaker-attributed transcripts from multi-speaker audio (interviews, meetings, podcasts, call-center recordings) where standard ASR produces a flat text stream without speaker labels.

Who should use it

researchers and product teams building multi-speaker transcription (meetings, podcasts, interviews, call-center)podcasters and journalists who need speaker-attributed transcripts without manual labelingdevelopers who want an end-to-end model instead of cascading ASR + diarization pipelinesEnglish + Mandarin bilingual teams (pretrained checkpoints ship for both)real-time applications (the Python SDK supports streaming inference)

Who should skip it

Skip if the source link, docs, or setup requirements do not match your workflow.

Risk explanation

250 stars and pushed 2026-06-04 — research-track, not a production-hardened SaaS; benchmark on your own audio before depending on it; Pretrained checkpoints cover English + Mandarin; other languages require fine-tuning on labeled data.

Evidence links

Closest alternatives / related signals

asrmulti-speakerspeaker-diarizationtranscriptionpytorchenglishmandarinpretrained