Hume AI alternatives: 5 options compared
The best Hume AI alternative depends on what you were using it for. Hume closed its expression measurement API on 14 June 2026, and its voice and text-to-speech APIs are scheduled to close on 13 November 2026. Developers who need a low-cost raw voice-emotion signal can look at Behavioral Signals. Researchers who need validated facial coding can look at iMotions (which includes Affectiva). Teams testing video ads can look at Adverteyes or Element Human. Teams who want an interpreted answer on a specific message, rather than raw scores to build on, can look at EchoDepth. No single product replaces everything Hume offered.
Disclosure: Cavefish makes EchoDepth, which appears in this comparison. We have tried to describe each option fairly, including where we are not the right choice. Details are based on public information as of October 2026 and change often, so verify current terms, pricing and availability with each vendor before you commit.
The options at a glance
| Option | Type | Best for | Main limitation |
|---|---|---|---|
| Behavioral Signals | Voice-emotion API | Developers needing a low-cost voice signal | Voice only; focus has moved toward security and deepfake detection |
| iMotions (incl. Affectiva) | Facial-coding research platform | Academic and lab research needing validated facial coding | Researcher toolkit; no voice or text analysis |
| Adverteyes (ex-Realeyes) | Ad testing with facial coding | Brand teams testing video ads | Built for ads rather than business communications |
| Element Human | Social-video testing | Testing creator and short-form content | Marketing-focused; limited public detail on pricing and scale |
| EchoDepth (Cavefish) | Multimodal message testing | An interpreted verdict on one high-stakes message | Not a raw per-minute developer signal; smaller vendor with limited independent reviews |
What changed, and why it matters
Hume AI was one of the best-known emotion AI companies, selling expression, voice and language measurement to developers through APIs. In 2026 it began withdrawing them. The expression measurement API closed on 14 June 2026, and the voice and text-to-speech APIs are scheduled to close on 13 November 2026. Reports also indicate that its technology has been licensed to Google DeepMind and that key team members have moved there.
If you built on Hume, the practical question is not who else sells emotion AI. It is what job your integration was doing. Hume sold raw scores to developers who then built something on top. Some alternatives sell the same kind of raw signal. Others sell a finished answer for a narrower job. Matching the replacement to the job is more reliable than matching it to the brand.
Each option: strengths and limitations
Behavioral Signals
A small voice-emotion and speech analytics company offering an API, with on-premise and air-gapped deployment options. It is the closest structural match to what Hume sold: a raw signal for developers.
- Low entry cost, with API pricing reported from around $0.02
- Secure deployment options, including on-premise
- Established deepfake voice detection if that is also a need
- Voice only, with no facial or text analysis
- Strategic focus appears to be shifting toward security and defence
- Limited UK presence and few public independent reviews
Best for: developers who need a cheap voice-emotion signal to build into their own product.
iMotions (including Affectiva)
Affectiva, the original facial-coding company, is now a module within the iMotions research platform owned by Smart Eye. It is widely used in academic and market research.
- Among the most thoroughly validated facial-coding science
- Reaches corporates and universities through research channels
- Backed by a listed parent company
- A research toolkit that typically needs hardware, skills and setup time
- Facial coding does not cover voice or language on its own
- Module pricing starts in the low thousands per module plus support
Best for: universities and insight teams who need validated, raw facial-coding data.
Adverteyes (spun out of Realeyes)
A London facial-coding ad-testing business spun out of Realeyes in late 2025, after Realeyes moved its main focus to identity verification.
- Large body of test-session data for benchmarking ads
- Results reported in under 24 hours
- Sold through established partners including Nielsen and Kantar
- Face only, and designed for advertising
- A new spin-out with limited public track record
- Pricing from roughly $4,000 per test is a poor fit for frequent, low-cost checks
Best for: brand and media teams testing video ads against benchmarks.
Element Human
A London company testing social and short-form video using facial coding and eye tracking, with clients that include major broadcasters and agencies.
- Strong fit for social and creator content
- UK-based with recognisable clients
- Combines face and attention measures
- Marketing-focused rather than general-purpose
- Face and eyes only, with no voice or language analysis
- Pricing and company news are not widely published
Best for: marketing teams testing short-form and social video.
EchoDepth (Cavefish)
EchoDepth tests how a specific message is likely to land by measuring real people's reactions across face, voice and language, then returns a plain-English verdict with the changes to make and an audit trail. It also has an API, but it returns interpreted results on a message, not a stream of raw scores.
- Combines face, voice and text in one result
- Interpreted verdict plus recommended changes, so teams do not have to read raw data
- Fixed-fee pilots with nothing to integrate, and on-premise options
- Not a cheap per-minute raw signal, so a poor fit if you are building your own product on top
- Smaller vendor, and independent public reviews are still limited
- Aimed at high-stakes messages rather than high-volume, low-cost scoring
Best for: teams who want an answer on a specific message, such as a regulatory letter, results call or restructure announcement.
How to choose a replacement
Start with the job. If your product consumed Hume's scores as an input, you need another raw-signal provider, and the main questions are modality (voice, face or both), price per minute, and where the data is processed. If a person was reading Hume's output to make a decision, an interpreted result may suit better than a like-for-like API.
Check continuity before price. Several vendors in this market are small, and the category has already seen one exit and one pivot. Ask any replacement for data export terms, on-premise or self-hosted options, a notice period for API changes, and what happens to your historical data. These questions are worth asking of us too.
Then check the regulatory scope. Tools aimed at audiences and customers are generally treated differently from tools that assess employees. The EU AI Act prohibits emotion recognition in workplaces and education, with narrow exceptions, so how and where a tool is used matters as much as which tool it is. Take legal advice for your own use.
When EchoDepth is not the right choice
If you need a low-cost raw signal to power your own application, a developer API such as Behavioral Signals is a better fit than we are. If you are running academic studies that depend on validated, raw facial-coding data, iMotions is built for that. If your only goal is benchmarking video ads against a large dataset, Adverteyes has more relevant data than we do.
EchoDepth is worth a conversation when the real need is knowing how a particular message will land before it goes out, with a record you can show a board, regulator or investor.
Frequently asked questions
What are the best alternatives to Hume AI?
There is no single replacement, because Hume sold raw emotion signals to developers and the alternatives each cover part of that. Behavioral Signals is the closest for voice-emotion APIs. iMotions, which includes Affectiva, is the established choice for validated facial coding in research. Adverteyes and Element Human focus on video and ad testing. EchoDepth offers an interpreted verdict on a specific message across face, voice and text. Match the alternative to the job your Hume integration was doing.
Is Hume AI shutting down?
Hume AI has closed its expression measurement API, which stopped on 14 June 2026, and its voice and text-to-speech APIs are scheduled to close on 13 November 2026. Reports also indicate that its technology has been licensed to Google DeepMind and that key team members have moved there. Check Hume's own announcements for the latest position on any product you use.
When do Hume AI's APIs close?
The expression measurement API closed on 14 June 2026. The voice and text-to-speech APIs are scheduled to close on 13 November 2026. If you depend on either, plan your migration and data export before those dates and confirm timings with Hume directly.
Is there a cheap Hume AI alternative for developers?
Behavioral Signals offers a voice-emotion API with pricing reported from around $0.02, with on-premise and air-gapped options. It covers voice only. For facial analysis, iMotions is a research platform with higher entry costs. Always compare accuracy, languages and data-handling terms as well as price.
Can EchoDepth replace Hume AI?
Only for some uses. EchoDepth returns an interpreted verdict on a specific message, with recommended changes and an audit trail, rather than a stream of raw per-minute scores. If you used Hume's raw scores as an input to your own product, a developer API is a closer match. If you used the output to judge how a communication would land, EchoDepth may be a fit.
Information based on public sources and vendor announcements, last reviewed 29 October 2026. Pricing and availability change, so verify with each vendor.
What are the best alternatives to Hume AI now that its emotion APIs have closed?
There is no single replacement for Hume AI, because it sold a developer-facing raw signal and the alternatives each cover a different part of that. Behavioral Signals is the closest for voice-emotion APIs, iMotions for validated facial coding in research, and Adverteyes or Element Human for video ad testing. EchoDepth suits teams who want an interpreted verdict on a specific high-stakes message, not raw scores.
Not sure which kind of replacement you need?
Tell us what your Hume integration did. If the right answer is a developer API, we will say so.
Talk it through