An audio and text annotation platform.
Annotators label, reviewers check, and managers accept the work.
Teams that trust us
- Mohirdev
- Data-Hub
LabelCraft in numbers
Figures from the production platform, mostly Uzbek and multilingual speech data.
- Hours of audio uploaded
- 2,100+
- Audio files uploaded
- 40,000+
- People on the platform
- 215+
- Annotation projects
- 20+
What goes wrong without a review process
- Mixed tasks
- When one person transcribes, tags and summarises a file in one sitting, a weakness in one skill affects every layer.
- No history
- In a shared spreadsheet a reviewer's fix overwrites the annotator's text, so nobody can later say whose mistake it was.
- Pay disputes
- Without a record of who wrote each version, a disagreement about pay cannot be settled from the data.
- Locked data
- With many vendors, getting your full dataset back takes a support request or an extra fee.
What you can label
Speech & audio
Transcription with speakers and timestamps, and a category for each segment.
- Speaker diarization
- Speech regions
- Sound events
- Audio classification
- Transcription
- Category review
Text & NLP
Named entities and summaries for the text you upload.
- Text classification
- Named entities
- Translation
- Summary
- Slots and intentsSoon
- Question answeringSoon
- Relation extractionSoon
- Entity linkingSoon
Conversations & LLM data
Dialogue labels, written and rated responses, preference pairs and rankings. These are in development.
- Dialogue turn labelsSoon
- Response writingSoon
- Response ratingSoon
- Side-by-side choiceSoon
- Preference pairsSoon
- Relevance rankingSoon
Tasks can follow each other. For example, a finished transcript can go on to entity tagging, with its own annotators and reviewers.
What the studio does
Start freeAI draft
AI can prepare the first transcript, entities or summary. The annotator corrects it instead of starting from an empty page.
Three-step review
An annotator does the work, a reviewer checks it and a manager accepts it. None of the three steps can be skipped.
Version history
Every save is a new version. You can compare any two versions segment by segment and restore an older one.
Analytics
Hours, quality scores and penalties for each annotator, reviewer, project and team.
How work moves through a project
- 01 · Upload
Add audio files, or text as .txt, .json or .jsonl, one at a time or all at once.
- 02 · Annotate
In the studio, the annotator transcribes the audio, splits it into segments and marks the speakers. In text, they mark entities and write a summary.
- 03 · Review
A reviewer approves the work or corrects it. Each review gets a quality score.
- 04 · Accept
A manager accepts the item. Accepted items make up your dataset and can be exported.
Your data
- Storage
- Each organization's files are kept in separate storage.
- Export
- You can export the full dataset yourself at any time, free of charge.
- API
- Accepted data is also available through the API and the MCP server, with access keys you control.
- Self-hosting
- With an Enterprise licence, the whole platform runs on your own servers.
Pricing by stored audio
Start freeShowing monthly prices.
Personal
For one person working alone, without reviewers.
- One member: your personal workspace
- Up to 5 hours of stored audio and 10,000 text items
- 120 AI credits, once — about 2 h of transcriptionOne-time, never monthly: AI credits are added once — as a gift when you sign up, and as a bonus with your first paid size — and never reset or expire. The monthly price pays for storage.
- Unlimited projects
- Full studio and personal analytics
- Community support
Team
or $79 a month with yearly billing
The price does not depend on team size
For a team with annotators, reviewers and a manager.
- Unlimited members, each with their own account
- Up to 250 hours of stored audio and 500,000 text items
- 4,200 AI credits, once — about 70 h of transcriptionOne-time, never monthly: AI credits are added once — as a gift when you sign up, and as a bonus with your first paid size — and never reset or expire. The monthly price pays for storage.
- AI drafts from your own model endpoint
- Unlimited projects
- Reviewer and manager check every item
- Quality scores, penalties and disputes
- Analytics with XLSX export
- Email support, answered the next business day
Scale
or $239 a month with yearly billing
For teams that send accepted data into their own training pipeline.
- Everything in Team
- Up to 700 hours of stored audio and 1,400,000 text items
- 13,200 AI credits, once — about 220 h of transcriptionOne-time, never monthly: AI credits are added once — as a gift when you sign up, and as a bonus with your first paid size — and never reset or expire. The monthly price pays for storage.
- AI cross-check of transcripts and spot checks
- External API, access keys and MCP
- Scheduled backups
- Priority support on the same business day, 99.5% uptime SLA
Enterprise
Self-hosted licence with support, billed yearly
For organizations whose data must stay on their own servers.
- On your own servers or a dedicated deployment
- Unlimited members, storage and AI
- Yearly licence with a support contract
- SSO and custom integrations
- Onboarding and team training
- Uptime SLA with a named support engineer
- Managed hosting on request
The monthly price pays for storage: the audio hours and text items you keep. AI credits come once and never expire; one credit covers one minute of transcription or about 20,000 characters of entity tagging. Team and Scale include unlimited members.
Frequently asked questions
What can we label?
Audio files, and text as .txt, .json or .jsonl. The task list above shows what is available now and what is in development.
Which languages can we annotate?
Any language your team writes. The interface itself is in Uzbek, Russian and English.
Can the AI drafts come from our own model?
Yes. On the Team plan and above you can connect your own model endpoint.
What counts as an hour of stored audio?
The total length of the audio kept in your workspace. A 40-minute file counts as 40 minutes whether it is annotated or not.
What happens when we reach the storage limit?
New uploads stop until you delete files or projects, or move to a larger size. Annotation, review and export keep working, and nothing is deleted.
Does archiving a project free storage?
No. Storage counts audio hours and text items, so only deleting files or a project frees room. Archiving hides a finished project and keeps everything in it.
Do we pay for adding team members?
No. You pay only for hours of stored audio, and every plan includes unlimited members. Each person needs their own account for penalties to be fair.
Can we run LabelCraft on our own servers?
Yes, with an Enterprise licence. You get unlimited members, storage and AI on your own infrastructure, with a support contract and an uptime SLA.
How does yearly billing work?
You pay for twelve months in advance, and each month costs 20% less.
Try LabelCraft on your own data
The personal plan is free. Team plans are priced by stored audio, with unlimited members.