Maybloc
Speech-to-text built for conversations with more than one voice.
About
Maybloc launched October 15, 2026 with a narrow thesis: transcription that survives overlapping speech. Under it is Calico, a speech model built on open research including BUT's DiCoW, which Maybloc says posts the lowest diarization (who-said-what) error on NOTSOFAR, AMI, and DiPCo against pyannote, AssemblyAI, and Deepgram — over 40% fewer errors than AssemblyAI and pyannote on NOTSOFAR's real office meetings. It keeps every speaker apart, keeps overlapping speech instead of dropping it, and timestamps every word. It covers 45 languages and up to 25 speakers, handles files up to 3 hours, and costs $0.006 per minute billed by the second, with a dashboard and REST API plus add-ons for subtitles, PII redaction, summaries, speaker names, and translation.
What it can do
- Multi-speaker transcription with word-level timestamps (provider-claimed)
- Lowest diarization error on NOTSOFAR, AMI, DiPCo vs pyannote, AssemblyAI, Deepgram (provider-claimed)
- Overlapping speech preserved, every speaker kept apart (provider-claimed)
- 45 languages, up to 25 speakers, files up to 3 hours (provider-claimed)
- Dashboard + REST API (provider-claimed)
- Add-ons: subtitles, PII redaction, summaries, speaker names, translation (provider-claimed)
Setup
- Sign up at maybloc.com.
- Upload audio (up to 3 hours) in the dashboard or send it to the REST API.
- Receive a word-timed transcript with speaker labels.
- Enable add-ons as needed: subtitles, PII redaction, summaries, speaker names, translation.
Prerequisites
Maybloc account; audio files up to 3 hours; $0.006/minute usage billing, billed by the second.
Use cases
Works with
Trust score — 5/10
Scored 7/10 (METRICS tier, metrics-v1). Evidence: launched Oct 15, 2026; detailed public launch post with benchmark tables and method link (maybloc.com/#benchmarks); builds on published open research (DiCoW from BUT); official URL verified live (200, Oct 2026); trending on Product Hunt Oct 16–17, 2026. Strengths: transparent methodology, commodity pricing, clear add-on list. Docked for: (1) all benchmarks provider-run with no independent replication; (2) brand-new company with no track record or third-party reviews; (3) closed model and API — claims are not auditable. License: proprietary API.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 3.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Launched Oct 15, 2026; speech-to-text focused on overlapping multi-speaker audio
- · Calico model built on open research including BUT's DiCoW; public benchmark tables + method page
- · Claims lowest diarization error on NOTSOFAR/AMI/DiPCo vs pyannote, AssemblyAI, Deepgram (provider-run)
- · 45 languages, 25 speakers, 3-hour files; $0.006/min billed by the second
- · manual review: all benchmarks provider-run with no independent replication; brand-new company with no track record; closed model/API — claims not auditable
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.