# Maybloc: Every voice, on the record.

Speech-to-text infrastructure for recordings. Every speaker separated, every overlap kept, every word timed, through a dashboard or a REST API.

## What you get back from every recording

- Every speaker, separated: each voice gets its own label and lane across the whole file.
- Crosstalk, preserved: when two people speak at once, both are transcribed and the overlap is marked with its exact start and end.
- Timed to the word: every word carries its own timestamp.
- Subtitles, ready to ship: export SRT or WebVTT with speaker names; the Subtitles add-on cuts them into caption-sized lines.

## Benchmarks

Maybloc has the lowest error rate of any system tested, on every benchmark. Full tables and how they were measured: https://maybloc.com/benchmarks.md (page: https://maybloc.com/benchmarks).

- 26% fewer errors than the next best system, on average across the public meeting benchmarks
- 43% fewer errors than AssemblyAI and pyannote on NOTSOFAR's real office meetings
- 54–73% fewer errors than Deepgram Nova-3
- Best published result on AMI headset meetings, ahead of Gemini 2.5 Pro, Gemini 3.0 Pro and VibeVoice

## Beyond the transcript

- PII redaction (+5 credits per minute)
- Smart formatting (free)
- Subtitles (free)
- Insights (free)
- Speaker names (+5 credits per minute)
- Sentiment (+5 credits per minute)
- Topics and entities (+5 credits per minute)
- Key terms (+5 credits per minute)
- Custom fields (+5 credits per minute)
- Translation (+15 credits per minute per language)
- 45 languages, or detected automatically.

## The model

Calico is the speech model behind every transcript, in the dashboard and through the API. Up to 25 speakers per recording, files up to 3 hours or 2 GB.

## Built into your product in an afternoon

A REST API at https://app.maybloc.com/api/v1, described by https://app.maybloc.com/openapi.json.

```sh
curl https://app.maybloc.com/api/v1/transcripts \
  -H "Authorization: Bearer $MAYBLOC_API_KEY" -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" -d '{ "upload_id": "UPLOAD_ID", "duration_ms": 155000 }'
```

```json
{ "speakers": [{ "id": "speaker_1", "speaking_ms": 61200 }], "turns": [{ "id": "t00001", "speaker_id": "speaker_1", "start_ms": 1200, "end_ms": 3400, "text": "...", "words": [] }], "overlaps": [{ "id": "o1", "start_ms": 3100, "end_ms": 3400, "turn_ids": ["t00001", "t00002"] }] }
```

## Your recordings stay yours

- Delete anytime: deletion is immediate and permanent.
- Private to your workspace, and never used to train models.
- Stored for playback: recordings for 7 days, results for 30 days.

## Pricing

- Audio: 60 credits per minute. $1 buys 10,000 credits; new accounts start with 10,000 credits.
- PII redaction: +5 credits per minute
- Smart formatting: free
- Subtitles: free
- Insights: free
- Speaker names: +5 credits per minute
- Sentiment: +5 credits per minute
- Topics and entities: +5 credits per minute
- Key terms: +5 credits per minute
- Custom fields: +5 credits per minute
- Translation: +15 credits per minute per language

Full map for agents: https://maybloc.com/llms.txt