Rev AI pricing

About

Rev AI is the developer/API arm of Rev.com, the long-running speech-to-text and captioning company. Where Rev.com sells human and automated transcription, captions, and subtitles directly to end users through a web app, Rev AI (rev.ai) exposes the underlying speech engines as pay-as-you-go APIs for developers — asynchronous and streaming automatic speech recognition (ASR), plus a suite of audio-intelligence features layered on top of the transcript.

The product line centers on Rev’s in-house Reverb ASR model family (Reverb, Reverb Turbo, and a multilingual Reverb Foreign Language model spanning 57 languages), alongside Whisper-based machine transcription (Whisper Fusion and Whisper Large) and the option to route audio to human transcriptionists through the same API. On top of any transcript, customers can call Insights add-ons — language identification, language translation, summarization, sentiment analysis, and topic extraction — each metered independently.

Rev.com was founded in 2010 (now headquartered in Austin, TX) and built one of the largest human-transcription marketplaces before pivoting hard into AI. That history is the durable asset behind the API: Rev’s in-house Reverb ASR was trained on 200,000+ hours of audio its own transcribers labeled — the largest human-transcribed corpus ever used to train an open-source model — which Rev open-sourced in October 2024 (arXiv 2410.03930) while pricing the hosted API at the same rates.

Rev AI competes with developer-first speech platforms such as Deepgram, AssemblyAI, and Speechmatics, as well as the speech APIs from OpenAI (Whisper) and the hyperscalers. Its differentiator is the combination of a self-trained ASR stack quoted at aggressive per-hour rates and a uniquely deep human-in-the-loop option ($1.99/min) reachable through the same interface — letting buyers blend cheap machine output with human-grade accuracy where it matters.

Pricing summary : How Rev AI’s per-hour and per-minute speech pricing works

Rev AI uses a pure pay-as-you-go (metered) model with no monthly subscription, no seat fee, and no minimum commitment — you pay only for audio processed, with every job rounded up to the nearest second and a 15-second minimum. Pricing spans three distinct billing units:

1. Per hour of audio (Reverb ASR): Reverb Turbo $0.10/hr, Reverb Transcription $0.20/hr, Reverb Foreign Language $0.30/hr. Rev’s own models are the only ones quoted per hour.
2. Per minute of audio: Whisper Fusion $0.005/min, Whisper Large $0.005/min, Human Transcription $1.99/min, Forced Alignment $0.003/min, Language Identification $0.003/min, plus Standard/Premium tiers for Language Translation ($0.002 / $0.025 per min) and Summarization ($0.002 / $0.025 per min).
3. Per 10 words of transcript (Insights): Sentiment Analysis $0.0008/10 words and Topic Extraction $0.0008/10 words — these meter on transcript length, not audio time.

A free allotment — “free credits equivalent to 5 hours of Reverb ASR,” usable across all products — starts every account, and a sales-led Enterprise plan replaces list rates with volume-based pricing.

What makes this different: Rev AI mixes three metering units (per hour, per minute, per 10 words) on a single pricing card and uniquely offers human transcription through the same pay-as-you-go API as its machine models, letting buyers trade cost against accuracy per file. This is closer to pure usage-based pricing than the seat-plus-usage hybrids common in SaaS, and its time-based meter aligns with the media-minutes billing pattern shared across speech vendors.

Pricing by product

Speech-to-Text — Reverb (Rev’s in-house ASR, billed per hour)

Speech-to-Text — Whisper & alignment (billed per minute)

Human Transcription (billed per minute)

Insights add-ons (audio intelligence)

Enterprise

Sales motions across products: PLG / self-serve for all pay-as-you-go ASR, Whisper, human transcription, and Insights add-ons; sales-led for Enterprise volume agreements.

Hidden costs : What a real Rev AI transcription bill adds up to

The cheap per-hour headline understates the bill once Insights add-ons and human review stack on top of raw ASR. Two real-world examples:

Podcast platform — 1,000 hours/month of English audio, Reverb + summaries

Topic Extraction’s per-10-words meter, not the per-hour ASR rate, dominates this bill — a reminder that transcript-length metering scales very differently from audio-time metering.

Compliance team — 200 hours/month, machine pass + human review on 10%

Routing just 10% of audio to human transcription multiplies the bill ~60x over the machine-only pass — the $1.99/min human rate is the single biggest cost lever in Rev AI’s catalog.

Want to estimate your own Rev AI bill? Use the Rev AI pricing calculator to model your monthly cost based on hours of audio, model choice, and add-ons.

Pricing evolution : From one $0.035/min rate to a per-hour Reverb catalog

Rev AI’s pricing page has moved through three eras: a single flat per-minute rate (2019–2021), a multi-unit per-minute catalog with human transcription and Insights add-ons (2022–2024), and the current per-hour Reverb structure (2024 onward). The machine-ASR rate fell from $0.035/min to an effective ~$0.0033/min over five years — a roughly 10× reduction — while the human rate moved the opposite direction, from $1.50 to $1.99/min.

Cadence

Tracked range: 2019-03–2026-06 (Wayback snapshots of rev.ai/pricing). Quarters not listed were verified stable (0 price changes, 0 SKU additions).

Notable changes

• 2019-03 to 2021-09 — A single Pay-as-you-go SKU at $0.035/min, first 5 hours free, “no contracts.” (Wayback rev.ai/pricing, 2019-03-22 through 2021-09.)
• 2021-10 — Enterprise tier surfaced a published anchor “Starting at $1.20/hr” and volume-discount messaging — Rev AI’s first on-page acknowledgement of a volume motion. (Wayback 2021-10.)
• 2022-06 — Machine Transcription cut to $0.023/min; Human Transcription added at $1.50/min through the same API; Insights add-ons (Language Identification $0.003/min, Sentiment Analysis and Topic Extraction $0.0008/10 words) launched; HIPAA compliance announced. (Wayback 2022-06; 2022-08.)
• 2023-10 — Machine Transcription cut to $0.02/min; Forced Alignment added; company HQ on the page moved from San Francisco to Austin, TX. (Wayback 2023-10.)
• 2024-09 — Language Translation and Summarization added, each with Standard ($0.002/min) and Premium ($0.025/min) model tiers — the last pre-Reverb price card. (Wayback 2024-09.)
• 2024-10-03 — Rev open-sourced the Reverb ASR and diarization models (paper arXiv 2410.03930, weights on HuggingFace as Revai/reverb-asr under a non-commercial license) and simultaneously repriced the API from per-minute to per-hour Reverb SKUs; Human Transcription rose to $1.99/min. (Wayback 2024-10-03/04; Rev blog: Introducing Reverb.)

The Reverb relaunch in detail

The October 2024 change is Rev AI’s most consequential pricing event and the one a buyer is most likely to misread. Three things happened at once:

1. The billing unit changed from minutes to hours. The old “Machine Transcription $0.02/min” SKU disappeared and was replaced by “Reverb Transcription $0.20/hr.” On a per-minute basis that is ~$0.0033/min — an ~83% cut — but the headline number got larger ($0.02 → $0.20), so a glance at the page reads as a price increase when it is actually a steep decrease. This is a textbook example of the value-metric framing problem: the same resource looks cheaper or pricier depending purely on the unit it is quoted in.

2. Rev anchored the price to its own open-source model. By open-sourcing Reverb (trained on 200,000+ hours of Rev.com’s human-transcribed audio) and publishing the exact same $0.20/hr and $0.10/hr rates on the API, Rev tied its hosted price to a model anyone can download. The non-commercial license is the wedge — self-hosters who want commercial use are pushed back toward the paid API or a commercial license, a usage-based monetization play layered on top of an open-weights release.

3. Human transcription went the other way. While machine rates fell, Human Transcription rose from $1.50 to $1.99/min — widening the machine-vs-human gap to roughly 60× and reinforcing human review as the premium, accuracy-critical SKU rather than a default.

What’s unique : Per-hour ASR, human-in-the-loop, and per-10-words intelligence

1. In-house ASR quoted per hour, not per minute. Rev AI is one of the few speech vendors to list its own models (Reverb, Reverb Turbo, Reverb Foreign Language) per hour of audio rather than per minute, with Whisper models still per minute on the same card. The per-hour framing makes the in-house models look dramatically cheaper at a glance ($0.20/hr ≈ $0.0033/min) and steers buyers toward Rev’s own engine. The unit switch arrived with the October 2024 Reverb launch — the prior “Machine Transcription $0.02/min” SKU became “Reverb Transcription $0.20/hr,” a ~83% cut that the larger headline number disguises (Wayback rev.ai/pricing, 2024-09 vs 2024-10).

2. Human transcription through the same pay-as-you-go API. Few API-first speech platforms let a developer route audio to a human transcriptionist via the same endpoint and the same billing surface. At $1.99/min it is roughly 600x the cheapest machine model, turning accuracy into an explicit, file-by-file purchasing decision.

3. Three metering units on one pricing card. Reverb bills per hour, Whisper/human/translation/summarization per minute, and Sentiment Analysis and Topic Extraction per 10 words of transcript. Mixing audio-time and transcript-length meters is rare and changes which line item dominates a bill depending on workload — a vivid case study in the value-metric problem that AI vendors keep wrestling with.

Strengths & weaknesses

Billing UX : Free credits, per-second rounding, and a collapsed rate card

• Free credits “equivalent to 5 hours of Reverb ASR” — a time-denominated free allotment that converts to credits usable across every Rev AI product, granted to new accounts with no credit card to start.
• Per-second rounding with a 15-second minimum — every transcription job is rounded up to the nearest second and billed for at least 15 seconds, so very short clips carry a floor cost.
• “View More Offerings” expander — the headline pricing card shows only the Pay As You Go and Enterprise plans; all per-unit rates are hidden until the user clicks “View More Offerings,” which expands the full Reverb / Whisper / Human / Insights rate list.
• Shared credit pool across products — free credits and (presumably) paid balance apply across ASR, human transcription, and Insights add-ons rather than being siloed per product.
• Enterprise “Let’s Talk” path — volume-based pricing, flexible commercial terms, a dedicated account manager, and higher data-control guarantees are quoted through a sales-led contact flow rather than self-serve.

Strategic wins : Decisions that strengthen Rev AI’s position

1. Open-sourcing Reverb to anchor the hosted price

In October 2024 Rev open-sourced the Reverb ASR and diarization models — trained on 200,000+ hours of its own human-transcribed audio (arXiv 2410.03930) — and published the API at the identical $0.20/hr and $0.10/hr rates. The non-commercial license means self-hosters who want commercial use are steered back to the paid API or a commercial license, turning an open-weights release into a funnel. Pairing this with the per-hour unit (vs Whisper’s $0.005/min) makes Rev’s own engine look like the obvious value choice — a classic value-metric framing move that nudges buyers toward the model Rev controls and can train cheaply.

2. Human transcription as a premium API SKU

Exposing $1.99/min human transcription through the same API keeps high-accuracy revenue inside the platform instead of losing it to manual workflows. It turns Rev’s legacy human-transcription business into an upsell that developers can call programmatically — a durable moat machine-only competitors like Deepgram lack, and a textbook example of usage-based pricing for SaaS and AI blending compute and labor in one meter.

3. Transparent public rates lower the funnel friction

Publishing every per-unit rate (once expanded) lets developers self-qualify and start on free credits without talking to sales, matching the self-serve, product-led motion that wins API budgets. The 5-hours-of-Reverb free allotment is a concrete, easy-to-grok trial offer — the kind of low-friction onboarding that usage-based pricing migrations are built around.

Areas to improve : Gaps in transparency and cost predictability

1. Stop hiding the rate card behind an expander

Burying every per-unit price behind “View More Offerings” hurts both buyers and AI-search visibility. Surfacing a comparison table by default — or at least the headline ASR rates — would reduce friction and improve how the page is cited by tools like Perplexity and AI Overviews.

2. Normalize the metering units

Mixing per-hour, per-minute, and per-10-words meters on one card makes back-of-envelope cost modeling error-prone. Publishing a single normalized ”$/hour-equivalent” column alongside each SKU (as some peers do for media-minutes billing) would make the catalog far easier to budget against.

3. Publish volume tiers instead of forcing Enterprise contact

The jump from public PAYG rates straight to “Let’s Talk” leaves mid-volume buyers with no self-serve discount path. A published volume-discount ladder would capture spend from teams too large for list rates but too small to negotiate.

Key takeaways

1. Unit choice is a pricing-strategy lever, not a detail. Rev AI quotes its own ASR per hour and Whisper per minute on the same card — the unit alone reframes which model looks cheapest. Pick the value metric that flatters the SKU you want to sell.
2. Human-in-the-loop can be a premium API SKU. Packaging $1.99/min human transcription behind the same endpoint as machine models keeps high-accuracy revenue on-platform and gives buyers an explicit cost/accuracy dial.
3. Free allotments read better when denominated in the customer’s unit. “5 hours of Reverb ASR” is more tangible than a dollar credit, even though it ultimately converts to credits.
4. Mixed metering units shift where the bill concentrates. Per-10-words add-ons (sentiment, topics) can dominate a transcript-heavy bill while per-hour ASR stays cheap — model the dominant meter, not the headline one.
5. Transparency only helps if it’s discoverable. Public rates hidden behind an expander forfeit much of the SEO and AI-citation benefit of publishing them at all.

UBP implications

1. Per-hour vs per-minute framing is a usage-pricing design choice. Quoting the same metered resource in a larger unit lowers the perceived price without changing the economics — a tactic any usage-based vendor can borrow.
2. Multiple meters on one product fragment the bill. When audio-time and transcript-length meters coexist, the dominant cost driver depends on the workload mix, complicating both forecasting and the vendor’s own revenue modeling.
3. Human labor as a metered SKU expands the usage-pricing surface. Rev AI shows that expensive human work can be sold per minute through the same API as compute, blending labor and inference into one consumption meter.

Sources

• Rev AI pricing page (accessed 2026-06-04)
• Rev AI API documentation (accessed 2026-06-04)
• Rev AI blog (accessed 2026-06-04)

Bottom line

Rev AI turns Rev.com’s speech engine into a transparent, pure pay-as-you-go API: in-house Reverb ASR from $0.10–$0.30 per hour, Whisper models at $0.005/min, and — uniquely — human transcription at $1.99/min through the same endpoint, with five per-minute or per-10-words intelligence add-ons on top. The catalog is cheap and flexible, but mixed billing units and a rate card hidden behind an expander make it harder to model and discover than it should be.

Want to compare Rev AI against other speech-to-text pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Live pricing page confirmed (Whisper SKUs visible)

Jun 2026

Live rev.ai/pricing capture: Reverb Transcription $0.20/hr, Reverb Turbo $0.10/hr, Reverb Foreign Language $0.30/hr; Whisper Fusion / Whisper Large $0.005/min; Human Transcription $1.99/min; Forced Alignment $0.003/min. Insights: Language Identification $0.003/min, Language Translation $0.002/$0.025/min, Summarization $0.002/$0.025/min, Sentiment Analysis $0.0008/10 words, Topic Extraction $0.0008/10 words. Free credits equal 5 hours of Reverb ASR; Enterprise volume-based, sales-led. Rate card hidden behind 'View More Offerings.'

Reverb open-source launch — ASR repriced per hour

Oct 2024

Wayback 2024-10-03/04 snapshots, coinciding with Rev open-sourcing the Reverb ASR + diarization models (arXiv 2410.03930, HuggingFace Revai/reverb-asr, non-commercial license): the per-minute 'Machine Transcription' SKU was replaced by per-hour Reverb SKUs — Reverb Transcription $0.20/hr, Reverb Turbo $0.10/hr, Reverb Foreign Language (57 languages) $0.30/hr. Human Transcription raised $1.50 → $1.99/min; Forced Alignment cut $0.02 → $0.003/min; free credits reframed to '5 hours of Reverb ASR.' This is the current rate structure.

Translation + Summarization add-ons (Standard/Premium tiers)

Sep 2024

Wayback 2024-09 snapshot: Language Translation and Summarization added, each with Standard ($0.002/min) and Premium ($0.025/min) tiers; an enterprise 'highest level of data control and security' line appears. Machine Transcription still $0.02/min, Human $1.50/min — the last pre-Reverb price card.

Machine ASR drops to $0.02/min; Forced Alignment added

Oct 2023

Wayback 2023-10 snapshot: Machine Transcription cut again to $0.02/min; Forced Alignment added at $0.02/min. Company HQ on the page moved from San Francisco to Austin, TX. Human Transcription held at $1.50/min.

Machine rate cut, human transcription + Insights added (multi-unit catalog)

Jun 2022

Wayback 2022-06 snapshot: Machine Transcription cut from $0.035 to $0.023/min; Human Transcription added at $1.50/min via the same API; Insights launched — Language Identification $0.003/min, Sentiment Analysis and Topic Extraction $0.0008/10 words. HIPAA-compliance banner appears; free credits reframed to '5 hours of Machine Transcription.' This is where Rev AI became a multi-unit (per-minute + per-10-words) catalog.

Enterprise gets a published anchor price

Oct 2021

Wayback 2021-10 snapshot: Pay-as-you-go held at $0.035/min, but the Enterprise tier swapped 'Custom pricing' for a visible anchor 'Starting at $1.20/hr' and the headline shifted to 'Commitment and no commitment options. Volume discounts available for enterprise.' First time a volume motion was surfaced on the page.

Single flat per-minute rate

Mar 2019

Earliest archived rev.ai/pricing snapshot (Wayback 2019-03-22): one Pay-as-you-go SKU at $0.035/minute with 'simple pricing, no contracts,' first 5 hours free, no usage limits, plus a Contact-us Enterprise tier with no published rate. By 2020-08 the same $0.035/min added 'rounded to nearest 15 seconds.'