Subtitling Rates per Minute: What to Charge When AI Drafts First

Published vendor rates for human-made translated subtitles run from about $6.50 to $30 per video minute, depending on language and seller. For an AI draft, quote by review time instead: measured review minutes per video minute times your hourly rate, plus a quality-check pass, tool cost and a per-file minimum.

No survey of freelance subtitling rates turned up when I looked. What exists is vendor price pages, which differ from one another by a factor of two to four for the same language, and forum threads that are anecdotal and often years old. Treat the market range as context, not a rate to copy. The number that is truly yours is the one you measure on your next three clips, and most of this tutorial is about measuring it.

Follow me on Instagram@sagnikteaches

What vendors actually print per video minute

Four vendor pages publish usable numbers. All are selling prices to a buyer, and each bundles timing, translation, a quality check and project management, so none of them is what a freelancer is paid.

Connect on LinkedInSagnik Bhattacharya

Read the spread before anything else. For Spanish, Rev lists $6.49 and GoLocalise says "from $24", about 3.7 times as much. For French, German, Italian and Portuguese the gap is $10.49 against $24, about 2.3 times. That is a difference in service level and sales model, a fixed price on a self-serve page against a starting price on a quote-led agency page, and not a mistake in either. It also means a translator who anchors on one page will be wrong by a factor of two or more about what "the market" charges.

Subscribe on YouTube@sagnikteaches

Price the review, not the draft

When an AI produces the first draft, the product you sell is still a finished, checked subtitle file. What changes is the labour behind it, so build the price from that labour. Nimdzi's 2020 analysis of text post-editing makes the point that effort, not edit distance, should set the discount, and that a fully polished result can cost as much effort as translating from scratch. A translator's post-editing workflow shows what that effort consists of. A flat "30 percent off because AI helped" is therefore a guess, in the client's favour or against you.

Price per video minute =
  ( (R + Q) / 60 x H + T ) x (1 + margin)

R      = review and edit minutes per video minute (measured)
Q      = quality-check minutes per video minute (second pass)
H      = your hourly rate
T      = tool cost per video minute
margin = 10 to 20 percent for admin, revisions and risk (your call)

Each input has a plain meaning. R is every minute spent correcting the transcript, fixing the translation and repairing timing and line breaks. Q is a separate pass, ideally after a break, with the video playing at normal speed. H is whatever hourly rate you need to earn, and nobody else can supply it. T is small: a subscription's AI top-up credits, such as Happy Scribe's $0.20 a minute, show how little the machine part costs next to the human part. The margin line pays for the revision round you will almost always be asked for.

Three clips, three prices

These figures are illustrative, built from the timing sheet in the next section. They show how the formula moves with difficulty and with your hourly rate, using T of $0.20 and a 15 percent margin.

Check the first row by hand: (4.0 + 1.5) / 60 x $60 is $5.50, plus $0.20 is $5.70, times 1.15 is $6.56 per video minute, so the 18-minute job quotes at about $118. At $60 an hour that lands within a few cents of Rev's Spanish tier, and at $80 it sits between Rev's first two tiers, while GoLocalise's "from $24" is a different product with different terms.

Clip C at $60 an hour comes to $10.24 per video minute, against Rev's $15.99 for Japanese. A vendor's price has to carry reviewers, project management and margin, so a freelancer sitting below it is plausible. A freelancer sitting far below it, as the $40 column does for a rarer language, is a prompt to check the hourly rate before the quote goes out. If your own sums disagree with the table, trust the formula and your timings.

Measure R on three clips before you quote anyone

Pick three clips of about three minutes each that represent your usual work: one clean, one with two speakers talking over each other, one dense with technical vocabulary. Run each through the same AI pipeline you will use for clients, then time every correction session with a stopwatch, split into three buckets.

Divide each total by three to get R: 12 / 3 is 4.0, 19 / 3 is 6.3 and 20 / 3 is 6.7. The sheet also tells you where the time goes, which is useful when the client asks why a noisy recording costs more. If transcript fixes dominate, the fix is a better audio-to-text step or a request for cleaner source audio. If translation fixes dominate, the fix is a glossary and a better engine for that language, which an AI-assisted workflow for small agencies sets up.

Repeat the exercise for the first two jobs of any new content type. R drifts with the client's accent, speed and vocabulary, and a quote built on last quarter's clean webinars will lose money on a panel discussion. Keep the sheet; after ten jobs you have a rate card based on your own evidence, which no vendor page can give you.

What the quality-check pass covers

Q is not a re-read of the text. It is a watch-through at normal speed, because most subtitle faults only show in motion. Netflix's timed text style guides are a widely copied reference even for non-Netflix work, and a few of their limits turn the pass into a checklist. The general requirements set a minimum of five-sixths of a second and a maximum of seven seconds per subtitle, and two lines at most. Its English guide allows 42 characters per line and up to 20 characters per second for adult programmes; its Swedish guide allows up to 17, so reading-speed limits move with the target language. Your client's own style sheet overrides all of this.

• Timing: each subtitle appears with the speech and clears before the next speaker; none hangs across a cut.
• Reading speed and line length: no line over the client's limit, no subtitle too brief to read, no awkward split between a noun and its article.
• Names, numbers and units: every figure checked against the audio, not the draft.
• Glossary and brand terms: product names left alone, agreed translations used every time.
• On-screen text: translated or covered where the client wants it.
• Technical file: correct format and encoding, file opens, no stray timecodes, final length matches the video.

Budget Q at 1.0 to 2.0 minutes per video minute until you have your own figures; the table above used 1.5 and 2.0. A second linguist checking someone else's work is different again, which is the next section.

The clean-up work an AI draft leaves behind

Clients who ask for a discount are picturing the draft arriving finished. These are the kinds of faults that fill R, with illustrative examples of how an English-to-Spanish draft goes wrong and what the corrected line looks like.

The dropped negation is the one to keep in mind. It reads fluently, passes a spelling check and reverses the meaning, which is why the review must be done against the audio and not against the draft. Checking AI translations before customers see them covers a repeatable routine for exactly this.

Light review, full review and a proofread: three products

A client who says "the AI does the first draft" is often really asking about the level of finish. Pricing three levels from three separate timings is cleaner than discounting one. The figures below are illustrative, at $60 an hour with the same $0.20 tool cost and 15 percent margin.

Nimdzi's 2020 piece put light post-editing discounts on text at up to 35 to 40 percent off a regular rate. Subtitles differ because timing and line breaks are extra work, so do not borrow that figure. Time a light review the way you timed a full one and let the number fall where it falls. There is also a fourth level to decline: AI draft with no human review. A client can buy that from a platform for cents a minute, and your name on unchecked machine output is the one risk the formula cannot price.

Proofreading uplift: when a second linguist is part of the price

Some clients, or some content, call for a second person to proofread the subtitles independently. The only published figure to anchor an uplift is Happy Scribe's listing of human proofreading from $2.00 a minute ($1.90 on its Business plan). Add that to clip A at $60 an hour and the price becomes $6.56 plus $2.00, or $8.56 per video minute, an uplift of about 30 percent.

That number is worth holding in mind for two reasons. First, "AI draft, your review, independent proofread" lands at roughly the price of the cheapest self-serve tier, so you need to be certain the proofreader is as good as your promise. Second, it gives you a clean sentence for a client who asks for double-checking: the second pair of eyes is a named line item, at about 30 percent, not a vague extra.

Turnaround, rush and minimum fees

Published turnaround is thin. Rev says translated subtitles arrive in 48 hours or less, and GoLocalise says only that standard scheduling is more economical than rush delivery. I found no page publishing a rush multiplier for subtitles, so any uplift has to be yours. As a starting suggestion, not market data: add 25 percent for next-day delivery, 50 percent for same-day and refuse same-day where it would squeeze the quality-check pass.

Minimum fees matter more than most freelancers expect. GoLocalise prints a minimum per file of $225 to $300. A 90-second social clip at $6.56 a minute is $9.84, which does not cover opening the file, converting it, delivering it and invoicing it. If that handling takes 20 minutes at $60 an hour, that is $20 before any review. A minimum of $35 to $45 per file is a sensible floor to test. Bundle short clips into one batch quote where you can, so the minimum is paid once.

The same logic applies to your own studio if you are a video agency adding subtitles to production work. The subtitle line on the rate card should show the per-minute price, the minimum and the turnaround it assumes, which fits a video production rate card without a separate negotiation.

A three-language bundle, priced from the same sheet

A small video agency delivering a 20-minute product video with Spanish, French and German subtitles shows the bundle case, using freelance reviewers at $50 an hour and the same $0.20 tool cost and 15 percent margin. The reviewers' timings, illustrative again, differ by language.

For context only, Rev's published tiers for the same three languages add up to $6.49 plus $10.49 plus $10.49, or $27.47 a minute and about $549 for 20 minutes. GoLocalise's "from" rates for these languages are $24 each, $1,440 before its per-file minimums. The AI-first bundle at about $396 sits roughly 28 percent below Rev's list and far below the agency's, and it only holds if the timings do. Build the English timed master once and translate from it, so the spotting is paid for once and the second and third languages should come in lighter than the sheet assumes.

Writing the quote so the maths survives

A per-minute figure with no terms invites the argument that AI made it free. State the workflow, what the price covers and what it does not.

Subtitles, English to Spanish, 18-minute training video
Workflow: AI-assisted first draft, fully reviewed and quality-checked by a linguist
Price: $6.56 per video minute = $118 for this file (minimum per file: $40)
Includes: one revision round within 5 working days of delivery, SRT and VTT files
Excludes: re-timing after the video is re-edited; second-language proofreading (adds $2.00 per video minute)
Turnaround: 3 working days from approved source file; next-day delivery adds 25 percent
Glossary: client terms used as supplied; changes after delivery are billed at the hourly rate

Send the glossary request with the quote, not after it. A client who supplies product names and preferred terms up front cuts your R on the first job and gives you an easy rule for later disputes.

Replying to "it's AI, so it should cost less"

This request will arrive. A reply that works keeps the price tied to the work and avoids defending the technology.

Client: Since AI does the first draft, can you drop the rate by half?

Reply: The price follows the time it takes a linguist to make the subtitles correct and
readable, not who types the first draft. On your last two clips that was about four
minutes of checking for every minute of video, plus a full watch-through. If a cleaner
source file or a supplied glossary shortens that, the price falls with it, and I'll
show you the timing sheet. Half the rate would mean skipping the quality-check pass,
which I'd rather not do on subtitles that carry your name.

Notice what the reply offers: a lower price when the measured effort is lower, and evidence. It does not concede a flat discount, and it names the exact step that would be sacrificed. Clients who see their own cleaner files reduce the price usually stop asking.

Check the maths after every job

A quote is only a forecast. After delivery, add up the minutes you actually spent, including the revision round, and work out the hourly rate the job really paid. The 18-minute Spanish job quoted at $118 makes a good example. Planned time was 18 x 5.5, which is 99 minutes. Suppose the client's revision round pushed the real total to 124 minutes. The tool cost was 18 x $0.20, or $3.60, so the job paid ($118 minus $3.60) divided by 2.07 hours, about $55 an hour against a $60 target, or 92 percent of it.

One job at 92 percent is noise. Two jobs of the same type below roughly 90 percent is a signal to act: raise R in the next quote, or narrow what the revision round covers, for instance to corrections of errors and not changes of wording. Add a "revision minutes" column to your timing sheet, because revisions are the usual hidden cost and the first thing a flat discount erases.

Keep one sheet per client with R, Q and revision minutes for each job, and review it after every ten files. If R drops steadily, share some of that saving as a loyalty rate. If it climbs, the engine or the source has changed and the next quote needs to move.

Pricing questions translators ask

What is a typical subtitling rate per minute for a freelancer?

No published survey turned up, so there is no honest single figure. Vendor pages for human-made translated subtitles list roughly $6.49 to $30 or more per video minute depending on language, and freelancers' asking rates vary around that. Use those pages as context, then derive your own rate from the review time you measure and the hourly rate you need.

Should AI-drafted subtitles cost less than subtitles made from scratch?

Only if your measured review time is lower, and by the amount of time saved rather than a flat percentage. A 2020 analysis of text post-editing from Nimdzi noted that a fully polished result can take as long as translating from scratch, and that edit distance is a poor guide to effort. Time three real clips before promising any discount.

Should I quote per video minute or per hour?

Quote per video minute with a per-file minimum, built from your hourly effort, because clients compare per-minute prices across vendors. Keep hourly billing for open-ended work such as revisions after the source video changes. Say in the quote what one revision round covers so the per-minute price does not quietly turn into unpaid hours.

Do I have to tell clients a subtitle draft was made with AI?

Check your contract and the client's own rules, because many clients ask and some forbid it. Stating the workflow in the quote ('AI-assisted first draft, fully reviewed and quality-checked by a linguist') avoids disputes later. This is practical advice, not legal advice; ask a qualified adviser if a contract is unclear.

Further reads

• How Translators Use AI to Keep Terminology Consistent — Keeps names and terms steady across a series of videos.
• Does Your Translation Business Need an AI Usage Policy? — Writes down when AI is used, so clients are told the same thing.
• How Videographers Use AI to Edit, Caption and Quote Faster — The video-side view of captions and quoting.
• memoQ vs Trados for Small Translation Agencies: AI Features and Cost — Tool choice for agencies that outgrow a spreadsheet and a subtitle editor.
• Virtual Assistant Pricing When AI Does the Drafting: Rates and Packages — Another worked case of pricing a service when AI does the first draft.
• Is It Safe to Put Client Documents Into AI Translation Tools? — Confidentiality checks before client footage goes into an AI tool.
• Grant Writer Rates: What to Charge When AI Drafts Faster — Freelance grant writer rates by hour and by project, plus a worked quote showing how AI drafting should change your pricing, and why percentage fees are out.
• MTPE Rates per Word: How Translators and Agencies Price Post-Editing — Published light and full MTPE rate bands with sources named and hedged, an effort-based day-rate sum, quality tiers and wording for the quote.
• AI Subtitle Generator for Client Videos: Accuracy and Delivery Checks — Measure an AI subtitle generator on your own audio, then run line-length, reading-speed, export and QC checks before a client video ships.
• AI Subtitle Translation for Agencies: Quality Checks Before Delivery — A cue-by-cue workflow for AI subtitle translation, with timing, line-length and meaning checks, a QC sample and a brief for the native reviewer.
• AI Video Translator for Corporate Video: Dubbing, Subtitles, Checks — Choose subtitles or dubbing, cost each language and review the translated video scene by scene.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Rev pricing page; GoLocalise subtitling rates page; GoTranscript subtitling page; Happy Scribe pricing pages; Netflix timed text style guides (general, English and Swedish); Nimdzi analysis of post-editing pricing (2020). Checked 1 and 4 November 2026.