Agencies bill transcription in more ways than "per audio minute." Some shops sell pure transcript deliverables. Others bury transcription inside video editing, podcast production, research coding, or subtitle packages. Getting the model right protects margin and sets client expectations when AI drafts cut your internal cost but not your liability for wrong words.
This guide covers common billing models and when each fits. For tooling, see best transcription software for agencies. File Transcribe fits agencies that upload client recordings, edit speakers, and export TXT, DOCX, SRT, or VTT without running a meeting bot.
Billing models at a glance
Per minute
- How you charge
- Rate x audio length
- Best when
- Pure transcript jobs
- Watch-outs
- Noisy audio and accents eat edit time
Per file / per deliverable
- How you charge
- Flat fee per recording or package
- Best when
- Predictable client budgets
- Watch-outs
- Scope the revision rounds
Retainer
- How you charge
- Monthly hours or package
- Best when
- Recurring podcasts or weekly meetings
- Watch-outs
- Define rollover and rush rules
Bundled in edit
- How you charge
- Transcription included in edit/subtitle fee
- Best when
- Video and podcast agencies
- Watch-outs
- Do not underprice silent cleanup work
1. Per-minute billing
Classic freelance and agency transcription pricing: quote a rate per audio minute (or per finished hour), sometimes with tiers for rush, verbatim, or multi-speaker.
Works well when: The deliverable is the transcript itself. The client understands that 60 minutes of audio is not 60 minutes of work if the audio is clean.
Breaks when: Source audio is terrible, speakers overlap constantly, or the client expects broadcast-ready captions at the same rate as a rough notes dump. Build surcharges for poor audio, heavy accents needing extra review, and true verbatim.
AI angle: Use AI to draft faster, but bill for the reviewed deliverable. Do not promise human accuracy at AI-only cost unless your contract says AI-only draft.
2. Per-file or per-deliverable pricing
Charge a flat fee per interview, per episode, or per subtitle package. Clients like round numbers. You like not arguing about whether intros count.
Works well when: Files are similar length and quality. You have history on how long cleanup takes.
Breaks when: A "one episode" suddenly includes three hours of bonus content. Write length caps into the SOW (for example, up to 45 minutes included; overage at a stated rate).
3. Retainer for recurring transcript work
Monthly fee for a set number of episodes, meetings, or subtitle hours. Common for podcast agencies and content studios.
Works well when: Volume is steady. You want predictable cash flow and the client wants predictable cost.
Breaks when: Rush weeks stack up with no rush clause. Define turnaround SLAs, what happens to unused hours, and how AI-assisted vs human-reviewed work is labeled on invoices if the client cares.
4. Bundled inside editing or subtitle packages
Many agencies never sell "transcription" as a line item. They sell a finished video with captions, a podcast with show notes, or a research report with coded quotes. Transcription is internal cost.
Works well when: Captions and notes are required to ship the creative work. Clients buy outcomes, not process.
Breaks when: You forget to price the cleanup time. AI drafts help, but someone still fixes names, timings, and SRT line breaks. Tooling that exports clean SRT/VTT (like File Transcribe) reduces that friction. Agency tool roundup: best transcription software for agencies.
How AI changes the quote (without confusing the client)
Practical patterns agencies use in 2026:
- AI draft + human review as the default SKU for most marketing and podcast work
- AI-only draft as a cheaper SKU with clear "not publication-ready" language
- Human transcription for legal-adjacent, PR-sensitive, or accessibility-critical jobs
- Internal use of upload tools on client files the agency already stores, not meeting bots on client calls unless the client approved them
File Transcribe supports the upload-first path: guest try for process tests, then paid plans for production volume. Start at the homepage or compare agency options in the agencies guide. Pricing page for plan context: /pricing.
What to put in the statement of work
- Source format accepted (MP4, WAV, M4A, link downloads)
- Deliverable formats (DOCX, TXT, SRT, VTT, PDF)
- Speaker labeling expectations
- Verbatim vs clean-read
- Turnaround and rush fees
- Revision rounds included
- AI vs human review level
- Confidentiality and file deletion timeline
- Who owns the transcript after delivery
Sample packaging ideas (illustrative, not a price list)
Draft notes
- Includes
- AI transcript, light speaker pass, DOCX
- Good fit
- Internal research dumps
Publish transcript
- Includes
- AI + human review, names pass, DOCX/PDF
- Good fit
- Blog and podcast show notes
Caption package
- Includes
- Reviewed SRT/VTT + timing pass
- Good fit
- YouTube and social video
Retainer studio
- Includes
- N episodes/month with fixed SLA
- Good fit
- Ongoing podcast clients
Set your own numbers from labor cost and tool cost. Do not copy a competitor rate blindly.
Margin traps to avoid
- Quoting per minute on cafe interviews without a poor-audio surcharge
- Including unlimited caption revisions in a flat edit fee
- Letting account managers promise same-day turnaround without a rush line
- Using meeting bots on client calls without client consent language
- Shipping raw AI text as "final" because the draft looked fluent
Fluent is not the same as correct. Build review time into every SKU that carries your name.
FAQ
Should agencies bill the same for AI and human transcripts?
Usually no. Separate SKUs, or state the review level on one SKU. Clients deserve clarity about who stands behind the words.
Is per-minute still fair when AI drafts are fast?
Yes if you are selling reviewed accuracy and formatting. AI cuts draft time; it does not remove responsibility for wrong names and broken captions.
Can we use File Transcribe on client files?
Yes for upload workflows your contracts allow. File Transcribe does not join client meetings as a bot. Upload recordings you already have permission to process.
Where do we compare tools for agency volume?
See best transcription software for agencies and /pricing.
What about meeting bots for agency client calls?
Only with client consent and policy fit. Many agencies prefer native record plus upload. See privacy risks of AI meeting bots.
Build the quote around the deliverable
Pick a billing model that matches what the client actually receives, then run the files through a tool your editors will use. Try a sample client recording on File Transcribe and export the formats you sell before you lock the SOW.
More guides
- Transcription guides
- Try File Transcribe free
- Transcript format guide
- Best transcription software
- Transcribe Zoom meetings
