Hume AI pricing

About

Hume AI is an empathic-AI company building voice and emotion-understanding models. Founded in 2021 by Alan Cowen — a psychologist and former Google AI researcher who led work on affective computing — Hume started with its Expression Measurement API (facial, vocal, and language emotion analysis) and expanded into voice with the Empathic Voice Interface (EVI), a real-time speech-to-speech model, and Octave, a text-to-speech LLM. The company raised a $12.7M Series A (January 2023, led by Union Square Ventures) and a $50M Series B (March 2024, led by EQT Ventures), for roughly $72.8M total, and reported 100,000+ developers and businesses using its APIs by late 2025.

Hume’s positioning is that voice AI should read and respond to emotional expression, not just transcribe words. That thesis drives a product line where the voice products (EVI, Octave) are sold as subscription tiers with usage overage, while the older expression-measurement capability stays a pure pay-as-you-go API.

For the most current information, visit Hume AI.

Pricing summary : How Hume AI’s pricing model works

Hume runs two billing systems side by side. The voice products — EVI (speech-to-speech) and Octave (text-to-speech) — share a seven-tier subscription ladder: Free ($0), Starter ($3), Creator ($7, $14 after a 50%-off first month), Pro ($70), Scale ($200), Business ($500), and custom Enterprise. Each tier bundles a monthly quota of Octave TTS characters (10K → 10M) and EVI minutes (5 → 12,500), plus rising limits on concurrent connections (1 → 30) and requests per minute (15 → 225). Once you exceed the quota, usage-based overage applies.

Overage rates improve as you climb: on EVI 3, overage falls from $0.07/min (Starter/Creator) to $0.06 (Pro), $0.05 (Scale), and $0.04 (Business); on Octave 1, TTS overage falls from $0.15 per 1,000 characters down to $0.12, $0.10, and $0.05. Higher tiers also unlock team seats (3 on Scale, 5 on Business), the commercial license, and voice conversion.

Hume also exposes a second, cheaper model in each product via an in-table toggle. EVI 4 MINI runs at roughly half the EVI 3 rate — $0.035/min (Starter/Creator), $0.03 (Pro), $0.025 (Scale), $0.02 (Business) — and Octave 2 roughly halves Octave 1’s per-character cost: $0.075/1,000 (Creator), $0.06 (Pro), $0.05 (Scale), $0.025 (Business). Octave 2 also stretches the same included character quota to about twice the minutes (e.g. 10,000 characters ≈ 20 minutes vs ≈ 10 on Octave 1). The bundled monthly quotas and tier prices are identical across models; only the per-unit overage rate and per-character efficiency differ.

Separately, the Expression Measurement API is pure pay-as-you-go with no subscription, metered per minute of media (video+audio, audio, or video), per image, and per word. As of this capture Hume’s billing docs route all rates to the unified pricing page, which lists the EVI/Octave tiers rather than per-modality expression rates; the historical per-modality rates (video+audio at the high end down to a fraction of a cent per word) are recorded under Pricing evolution. Enterprise adds volume discounts and SOC 2 Type II / GDPR / HIPAA compliance.

What makes this different: Hume meters voice the way buyers actually experience it — minutes of conversation and characters of speech — and then discounts the unit rate as commitment rises, so the subscription tier doubles as a volume-discount lever rather than just unlocking features. The pay-as-you-go expression API sitting outside the subscription ladder reflects its origin as Hume’s first product.

Pricing by product

The overage rates above are for the default models (EVI 3 for speech-to-speech, Octave 1 for text-to-speech). The pricing page exposes a model toggle in each product row, and the newer models are cheaper per unit at every tier:

Per-model overage rates (EVI 3 vs EVI 4 MINI)

Per-model overage rates (Octave 1 vs Octave 2)

Bundled monthly quotas, tier prices, RPM caps, and connection limits are identical regardless of which model you pick; only the per-minute / per-character overage rate (and Octave 2’s roughly-2× minutes-per-character efficiency) changes.

Sales motions across products: self-serve PLG for Free through Business (instant signup, no card on Free), pay-as-you-go for the expression API, and sales-led Enterprise. A 50%-off first month applies to the Creator tier.

Hidden costs : What Hume AI users actually pay

The subscription price is only the floor. Real Hume bills are driven by EVI minutes and Octave characters consumed beyond the bundled quota, each metered as overage at the tier’s rate. Because EVI conversations are billed per minute of speech-to-speech, a moderately used voice agent can blow through the included minutes quickly — a Pro plan’s 1,200 included minutes is only ~20 hours of conversation per month.

Other things to budget for: the expression-measurement API is entirely separate and adds to the bill on top of any voice subscription; team seats are gated to Scale ($200, 3 seats) and Business ($500, 5 seats); and compliance (SOC 2 Type II, GDPR, HIPAA) plus the unlimited-everything tier are Enterprise-only. Concurrent-connection caps (1–30) can also force an upgrade well before raw usage limits bite for apps with bursty traffic.

Want to estimate your own Hume AI bill? Use the Hume AI pricing calculator to model your costs based on usage patterns.

Pricing evolution : Hume AI pricing history and changes

Cadence

Tracked range: 2023–present, via the Series A/B announcements, EVI generation releases, a 2025-02-28 Wayback snapshot, and live 2026-06-09 and 2026-06-30 captures.

Notable changes

• 2023-01 — Series A ($12.7M, USV). Commercial product is the pay-as-you-go Expression Measurement API.
• 2024-03 — EVI 1 launches with the $50M Series B (EQT Ventures); voice billed per minute at ~$0.102/min.
• 2024-09 — EVI 2 cuts per-minute pricing ~30% to ~$0.072/min.
• 2025-02 — Pricing page shows Octave TTS plus a bundled-quota subscription structure (EVI 1/EVI 2 minutes + TTS characters per tier).
• 2026-06-09 — EVI 3 and Octave 1 sit on a seven-tier ladder; EVI overage runs $0.07→$0.04/min and Octave TTS overage $0.15→$0.05 per 1K characters; expression APIs stay separate pay-as-you-go.
• 2026-06-30 — Hume adds a cheaper-model toggle in each voice row: EVI 4 MINI at roughly half the EVI 3 per-minute overage ($0.035→$0.02/min) and Octave 2 at roughly half Octave 1’s per-character overage ($0.075→$0.025 per 1K, ~2× minutes per character). Headline tier prices and bundled quotas are unchanged, so this is effectively a quiet price cut for buyers who opt into the newer models. The Expression Measurement API’s per-modality rates were dropped from the live pricing page in the same update (recorded as historical below).

Historical Expression Measurement API rates

When the Expression Measurement API was sold as a standalone pay-as-you-go product, Hume published per-modality rates directly. As of the 2026-06-30 capture these are no longer shown on the live pricing page (which now lists only the EVI/Octave subscription tiers), so they are recorded here for reference:

Enterprise customers historically received volume discounts on every modality. Buyers needing current expression-measurement rates now have to request them through sales rather than read them off the page — a transparency regression relative to the fully-published voice tiers.

What’s unique : Hume AI’s distinctive pricing mechanics

1. Tier as a volume-discount lever. Unlike most freemium tools where the tier only unlocks features, Hume’s tiers also lower the per-unit overage rate — EVI drops from $0.07 to $0.04/min and Octave from $0.15 to $0.05 per 1K characters as you climb. The subscription effectively bundles a committed volume discount.

2. Two billing systems for two product eras. The voice products (EVI, Octave) are subscription + overage, while the original Expression Measurement API stays pure pay-as-you-go. Buyers can consume emotion analysis with zero subscription, a relic of Hume’s API-first origins that’s unusual to keep alongside a tiered SaaS ladder.

3. Per-minute voice metering with falling generational prices. EVI is billed by minutes of conversation, and each model generation has cut the rate (EVI 1 $0.102 → EVI 2 $0.072 → tiered $0.07–$0.04). Pricing tracks model-efficiency gains directly, passing inference savings to buyers.

4. A model toggle that doubles as a self-serve discount. Since 2026-06-30 each voice row carries a model switch — EVI 3 vs EVI 4 MINI, Octave 1 vs Octave 2 — and the newer model sits at roughly half the per-unit overage at every tier (Octave 2 also stretching the same bundled quota to ~2× the minutes). Rather than re-pricing the headline tiers, Hume layers the price cut behind a model choice, letting cost-sensitive buyers opt into ~50%-cheaper inference while leaving sticker prices and quotas untouched. The cut is real but only captured by buyers who notice and flip the toggle.

Strengths & weaknesses

Billing UX : Hume AI billing controls and transparency

• Billing controls — Self-serve signup and tier upgrades from Free through Business; no credit card required on the Free plan; a 50%-off first month on the Creator tier. Enterprise is sales-quoted.
• Usage visibility — The pricing page publishes every included quota, overage rate, RPM cap, and concurrent-connection limit per tier, giving strong forward cost visibility. An in-table model toggle (EVI 3 / EVI 4 MINI, Octave 1 / Octave 2) lets buyers compare per-unit overage rates between the default and the cheaper newer models before committing. Expression-measurement rates are listed per modality.
• Payment options — Card-based self-serve checkout for Free through Business; Enterprise is invoiced under a custom contract with volume discounts. Support is Discord on self-serve tiers and Slack on Enterprise.

Strategic wins : Why Hume AI’s pricing decisions worked

1. Metering voice the way buyers experience it

Billing EVI by the minute and Octave by the character maps pricing to units developers already reason about — conversation length and speech volume — rather than opaque tokens. See choosing the right usage metric.

2. Using tiers as a volume-discount engine

By lowering the overage rate as the subscription rises, Hume turned its tier ladder into a commitment-for-discount mechanism, nudging growing customers upward without a hard sales touch. Related: how AI companies structure pricing.

3. Passing model efficiency to price

Each EVI generation cut the per-minute rate (EVI 1 $0.102 → EVI 2 $0.072 → $0.04–$0.07 tiers), and the 2026-06-30 EVI 4 MINI / Octave 2 toggle dropped overage another ~50% at every tier. Visibly passing inference savings to buyers builds trust in a category where cost deflation is the story. Related: outcome-based pricing trends.

4. Cutting price without re-pricing the menu

Rather than re-cut the published tier prices, Hume’s 2026-06-30 change layered a ~50%-cheaper option behind a per-product model toggle. This lets it advertise stable headline prices while quietly extending a steep discount to cost-sensitive buyers who opt into EVI 4 MINI / Octave 2 — protecting revenue from buyers who never switch while still answering competitive price pressure. See how AI companies structure pricing.

Areas to improve : Gaps in Hume AI’s pricing approach

1. Reconcile the two billing systems

Running subscription-plus-overage voice alongside a pure pay-as-you-go expression API forces buyers to track two cost models. A unified view (or shared credits) would reduce confusion. See bill shock and cost unpredictability.

2. Per-minute EVI can surprise at scale

Voice agents that talk a lot burn included minutes fast, and overage at $0.04–$0.07/min adds up quickly. Clearer in-product spend alerts and projected-overage warnings would help buyers avoid surprises.

3. Connection caps decoupled from usage value

Concurrent-connection limits (1–30) can force a tier upgrade for bursty apps that aren’t otherwise near their minute/character quotas — a friction point worth softening with add-on connection packs.

4. Surface the cheaper model, and re-publish expression rates

The ~50% EVI 4 MINI / Octave 2 discount only reaches buyers who notice and flip the in-table model toggle; defaulting the table to the cheaper model (or showing both rates side by side in the plan picker) would let the cut actually land instead of rewarding only attentive readers. Separately, pulling the Expression Measurement API’s per-modality rates off the live page makes the once-PLG product effectively sales-quoted — re-publishing them would restore the forward cost visibility the voice tiers still offer. See bill shock and cost unpredictability.

Key takeaways

1. Meter on the unit buyers count. Hume bills voice in minutes and speech in characters, mapping price to lived experience rather than tokens.
2. Tiers can be a volume-discount lever, not just a feature gate. Hume’s overage rate falls as the subscription rises, rewarding commitment automatically.
3. Pass model savings to price visibly. Each EVI generation cut the per-minute rate, reinforcing trust as inference costs fall.
4. A cheaper model can be a price cut in disguise. Hume’s 2026-06-30 EVI 4 MINI / Octave 2 toggle dropped per-unit overage ~50% without touching sticker prices — a way to answer price pressure while protecting revenue from buyers who never switch.
5. Watch the secondary meters. Connection caps and seat gates can drive upgrades independent of headline usage, shaping real bills more than the sticker price.

UBP implications

1. Hybrid subscription + overage suits per-minute AI products. A bundled quota gives a predictable floor while overage scales with heavy voice usage — a clean fit for conversational AI.
2. Declining unit rates by tier align incentives. Lowering the per-unit price as customers grow encourages expansion without a renegotiation, a tactic UBP practitioners can borrow. The model-toggle discount adds a second axis: a cheaper SKU behind a self-serve switch lets vendors pass inference savings without re-pricing the published menu. See usage-based pricing strategy.
3. Separate value metrics for separate products is fine — if each is intuitive. Minutes for voice, characters for TTS, and per-word/per-image for expression each map to their product’s natural unit.

Sources

• Hume AI pricing page — live capture (accessed 2026-07-23)
• Hume raises $50M Series B and releases EVI (accessed 2026-06-09)
• Hume AI business breakdown — Contrary Research (accessed 2026-06-09)
• Announcing the EVI 3 API (accessed 2026-06-09)
• VentureBeat — Hume EVI 2 and per-minute pricing (accessed 2026-06-09)
• Wayback Machine — hume.ai/pricing, 2025-02-28 snapshot (accessed 2026-06-09)

Bottom line

Hume AI is an empathic-AI company (founded 2021, ~$72.8M raised through a 2024 $50M Series B) that sells real-time empathic voice (EVI), Octave text-to-speech, and emotion-measurement APIs. Voice is priced on a seven-tier subscription ladder from Free to $500/month Business (plus custom Enterprise), each bundling EVI minutes and Octave characters before usage-based overage — EVI overage falls from $0.07 to $0.04/min and Octave TTS from $0.15 to $0.05 per 1K characters as tiers rise. Its older Expression Measurement API stays pure pay-as-you-go, metered per minute, image, and word (per-modality rates historically ran from a high of video+audio per minute down to a fraction of a cent per word, and are no longer published on the live pricing page). A 2026 update also added cheaper EVI 4 MINI and Octave 2 model options at roughly half the per-unit overage of EVI 3 / Octave 1. Browse the pricing blueprint for more fully-researched company profiles.

Want to compare Hume AI against other voice and conversational AI companies? 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.

EVI 4 MINI + Octave 2 cheaper-model toggle

Jun 2026

Pricing page now exposes a per-product model toggle: EVI 4 MINI runs at roughly half the EVI 3 per-minute overage ($0.035→$0.02/min vs $0.07→$0.04), and Octave 2 roughly halves Octave 1's per-character overage ($0.075→$0.025 per 1K vs $0.15→$0.05) while ~2× the minutes per character — an effective ~50% overage cut for buyers who opt in. Headline tier prices ($0/$3/$7/$70/$200/$500/Custom) and bundled quotas are unchanged. The Expression Measurement API's per-modality rates were pulled from the live page in the same update.

EVI 3 / EVI 4 MINI + Octave 2, seven-tier structure

Jun 2026

Current structure: Free, Starter $3, Creator $7–$14, Pro $70, Scale $200, Business $500, Enterprise custom. EVI 3/EVI 4 MINI overage runs $0.07→$0.04/min; Octave 1/2 TTS overage $0.15→$0.05 per 1K chars. Expression APIs remain separate pay-as-you-go.

Octave TTS + multi-tier subscription structure

Feb 2025

Pricing page shows tiered subscriptions bundling Octave text-to-speech characters and EVI 1/EVI 2 minutes (Free through high-volume tiers), moving from pure API metering toward subscription + overage.

EVI 2 — ~30% per-minute price cut

Sep 2024

EVI 2 ships with per-minute pricing reduced to ~$0.072/min, roughly a 30% cut versus EVI 1's $0.102/min, while adding faster, more customizable voice.

EVI launch + $50M Series B

Mar 2024

Hume launches the Empathic Voice Interface (EVI 1) alongside a $50M Series B led by EQT Ventures. EVI is a speech-to-speech voice model billed per minute (EVI 1 ~$0.102/min).

Series A and early expression-measurement API

Jan 2023

Hume AI (founded 2021 by Alan Cowen) raises a $12.7M Series A led by Union Square Ventures. Commercial product at this stage is the pay-as-you-go Expression Measurement API for facial, vocal, and language emotion analysis.