Sunday, February 28, 2010

Product Review: Humintell's MiX 2

This in-depth review is the first of a series that aims to evaluate Humintell products. The product covered is named MiX 2, an advanced microexpression recognition training software. The review covers the main components of the product with an outline of its high and low points. This review differs from most reviews because I wish to share my appreciation, opinions and suggestions that can be useful to potential buyers, actual customers and Humintell itself.

User Interface

The user interface is solid, lean and clean. Simplicity is a strong asset when done right and Humintell understood that. There is no major hurdles or ambiguities that would prevent a smooth usage because the user interface is straightforward. There is no online help since it is straightforward but instructions are given at every appropriate moment. As a professional, I don't want to waste time around a complex GUI or CLI and the MiX 2 user interface matches my requirement.

There are many similarities with METT/SETT 1 user interfaces but the current combination of look-and-feel plus the actual content and usability makes MiX 2 feels like an entire new generation.

My impression is that Humintell products run on LAMP (Linux, Apache, MySQL and PHP) softwares and use AJAX technology. It cannot get more simple and reliable. The user interface responsiveness is better than the stand-alone applications METT/SETT 1, which are both developed in Adobe Director, but is still a bit sluggish. The user interface opens in a 1024x768 pop-up window and works great with Mozilla Firefox.

The only configurable parameter is the speed of expressions, available in Set Speed pane. It is possible to set the speed from a macroexpression to a microexpression, or otherwise faster or slower. Limit speed selection is checked by default and there is no way to uncheck it effectively; it is an oddity.

The simplicity is a great asset in MiX 2 user interface but it's not without few more oddities. The videos refer to “happiness” but the user interface displays “joy”, which is an atypical inconsistency from a leading expert in the field of emotions. The navigation is straightforward but the menu button ambiguously serves as both menu and back button, it's not a major issue except that there is no explicit reference on where the user is in the context (e.g. main > instruction videos > anger). Usage time is written in a counter-intuitive manner, where 0 h 35 min is expressed as 1 h – 25 min. It is also slightly inaccurate because it does not take into account periods of inactivity. Some buttons disappear while loading the next image. The image turn black when you repeat an expression, which is already neutral. The buttons repeat, flash and next are organized in a visually unappealing manner, where it feels that a button is missing. It's also my first time with red “done” button, it's very unconventional. Nothing overly significant here but I do notice every detail.

Instructional Videos

There are 16 fantastic instructional videos available in a discernible format (video: flash video, 780x348, 14 FPS; audio: mp3, 64 kbps, 22050 Hz), each has a total length of 00:00:32 and a size of approximately 3 Mb. The 32 seconds sequences are composed of an emotion or two shown in frontal and profile views and repeated 3 times in slow motion. The voice over is a classic deep tone and assertive voice, which is a great choice.

The videos are the highlight of the software and we cannot get enough of these. I have watched videos over and over but nothing comes close to these instructional videos, when it comes to learning. It's much harder to sharpen your teeth on real-life situation videos with lesser quality (lighting, setup, angles, image, blur in slow motion, moving subjects, camera moving around, etc).

It would have been great to have a set of videos from at least one actor expressing every emotion. Unfortunately, there is no video with the 3/4th view and it is a big let down. Videos depicting real-time microexpressions would be a great addition to the existing video library. Full screen videos is an overly missing piece and downloadable media would be a huge plus.

Some videos are showing two emotions, that is, two frontal and two profile views. The order is frontal views first and then profile views second. I feel that it would have been better if a frontal view would have been associated with its profile, and even 3/4th views, to let us associate the person and its views as a group and at once in our memory. This would allow us to create a really complete representation of the given emotion by the actor.

For the videos showing only one emotion, I wish the images would have been a bit closer to allow the very same mental association to take place. It would have been a great addition to provide the 3/4th view on the single emotion videos.

Pre and Post Tests

The pre test has the trivial purpose to assess and keep on record the ability of a user before the training begins. It's a great way to see the real improvement before-and-after the training. There is no limit to take the post-test but your pre-test is one-time only. Taking the post test will allow the user to assess his or her proficiency and earn a level, from novice to expert I suppose.

The speed of flashing expressions is relatively unreliable because the software loads images in real-time and heavily rely on asynchronous responses from the server, both causing noticeable lag. I did feel variable speed during my post-test and it became distracting. This is probably a generalized issue but caching images could eventually solve this.

You will receive a “certificate of completion” once you have completed the post-test with a proficiency level of expert, which I earned on my first run.

The problem with the certificate of completion is that it has no trading value whatsoever. It's basically a gimmick. Humintell should instead consider developing a certification programme, in the same spirit of FACS Final Test or even Reid Institute's certification programme. The certificate of completion could be readily replaced by an incentive to buy more products, such as a coupon, or eligibility for a contest, etc.

Completing any test will unveil statistics about your performance. This is a great tool to understand your weaknesses and work on them. There is only two set of statistics available at all time: results from the pre-test and current results from the post-test, and accessible only once completed a test. I am inclined to believe that the software does not take advantage of these results in training or during the tests, where it could increase the probability a given emotion, that is typically problematic for the trainee, to be selected and displayed. Moreover, and unfortunately, there is no way to practice a user-selected set of emotions.

Further statistics are not necessary to me because I have long standing experience in non-verbal communication. But I can easily imagine that any professional or devoted individual would enjoy to have a simple performance chart to evaluate their own performances along their training. The factors of interests are percentile representing the identification of each emotion, at each given speed, over a certain number of trials and/or time. This is also a great opportunity for Humintell to collect data on real users, learning the real deal.


Practising with MiX 2 is wonderful in many ways because it brings considerable improvements over its predecessor, METT 1. The two additional views (profile and 3/4th) are a home run in my book. There is more emphasis on diversity than before and it makes MiX 2 a rich training environment. The quality of the content is no less than irreproachable with high quality pictures and interesting variations in expressions (e.g. anger with open mouth), and a strong selection of actors.

The images are available in an acceptable format (jpeg, 360x360 px, 24 bit) and an approximate filesize of 130Kb. My favourite format would be something that is closer to reality, i.e. the actual size of a face, full screen or at least something along the line of SETT 1.

The practice and tests images should ideally contain a sequence of four (4) frames rather than two (2). The sequence should contain at least a neutral expression, transition to a microexpression, a microexpression, transition back to neutral expression, or more frames. The reason is because the current sequences are overly unnatural since microexpressions are not like on/off switches. Microexpressions are progressive and fluid even though they happen in a split second. It would be possible to determine earlier or more accurately microexpressions with a smoother sequence during training; this would be even more meaningful to those who also have a subtle expression training.

The sequences of images seem to auto-center and it is a distraction because it is not how we look at people in our daily lives. Auto-centering has however some advantages since it makes sure that we always see the face in the best possible conditions.

The greater diversity in race and gender is an interesting addition but the emphasis on it comes at a cost. There is an apparent oversight in that non-verbal communication and other behavioural sciences are based on the analysis of patterns and interacting elements. In a nutshell, it is more important in a training environment to have a single actor displaying the entire set of emotions than having an actor for each emotion. The diversification is second in priority because it leans on the ability to distinguish the same emotions from one actor to another; failing to provide a complete set of emotions for each actor significantly undermines the value of diversification.

I would like to provide a valuable customer feedback to Humintell and a clear message of what I expect in a microexpression training software. I want to train on a software that offers a series of actors expressing every universal emotion, in every available view. A set of 3-5 actors would be fine. This is a tremendous value for us because we can retrain over and over without the ability to memorize the content by associating a face with a specific emotion.

It would be incredible to display the FACS scores along each and every practice images. We've got to take our chance while we have Dr. David Matsumoto around. This would be a great value to those who are working to reach a master level, beyond emotions.

End-User License Agreement (EULA)

The End-User License Agreement is questionable in many ways and one of the most prominent downside of the services offered by Humintell. Few highlights are presented in the following paragraphs that are areas of concern.

The scope of use clause is particularly interesting for its obvious ambiguities. For example, a registered user is prohibited to download content (image, texts, data, etc.) from Humintell. The normal usage systematically violates this clause, for the simple reason that it is a service provided over a network. Content will be downloaded by your web browser. Why is it wrong in the first place anyway? As a law abiding citizen, I have decided against including any screenshots with this review – just in case.

The termination clause is equally shocking because it gives unilateral authority to Humintell to terminate your account at any time, with or without any reason, with or without prior notice. Clauses 6 through 11 will be still effective once the termination has been called upon. There would be no such issue with a stand-alone software.

The proprietary information clause exposes an idea that is not appropriate for customers but instead for employees. It is often seen in an employment contract, where any material is handed over before leaving the company in order to make sure that Intellectual Properties (IP) are protected.

Furthermore, it seems that it may infer an anti-competitive restriction as a by-product of the current wording. The software business is straightforward in this respect: every vendor buys softwares from their competitors, try them out, use similar ideas but no content, etc. The proper manner to proceed is to obtain a patent or otherwise applicable rights.

The most considerable element when it comes to both clauses, scope of use and proprietary information, is that I use everything I learn to any ends and any extends. I do not want a vendor tell me off on this one because it's entirely unreasonable. Prohibiting registered users to resell their services is however reasonable.

The no warranties clause weaves any right you have to get what you have paid for, if under delivered. Whether the service fails or they have mixed up happiness and sadness images with each other, you may ask but Humintell is not legally required to fix any of this. I do understand the necessity of limiting liability but the sole “AS IS” softwares that customers should tolerate are generally covered by an open source license and absolutely free-of-charge.

This End-User License Agreement is not professional and corporate friendly. The restrictions are simply too overwhelming and raise some questions that should be answered by an attorney. There are too many fine lines that may prevent normal usage in such environment, including training staff even without the software and content, and possibly without consideration whether the instructor has prior knowledge. Home users may consider the license annoying but tolerable.

My experience with Humintell leads me to believe that they mean well and will honour any reasonable request. They are simply willing to legally protect themselves because running a business is much harder and risky than one can think. Humintell has experience to earn as a software developer/vendor and a tip or two could help. How about reviewing your EULA, offering a limited warranty and get a proper liability insurance?

Public Relation and Customer Support

It is noteworthy to write about Humintell's attitude toward the general public, regardless of whether people are/will be customers or not. I believe they have a genuine interest to build a community beyond selling products. They are dynamic, warm and approachable. Dr. David Matsumoto runs his own blog and he is responsive to the public questions and comments; and, it's not every day we can get a hold on such leading expert. Humintell has also its own facebook page with samples, related articles and links, and contests.

The customer support is outstanding in every aspect; it is professional, respectful and steadfast. It has a spotless record with me so far and its future is bright. Humintell are interested in their customers and care about them.


The software market has changed a lot the last ten years and it is greatly affected by the economy and the competing non-software markets from entertainment to whatever else is available to buy. For this reason, the most compelling question is always “Does this product worth the price?

  • Quality content, images and videos
  • Improvements over its predecessor
  • Best offer on the current market
  • Simple highly functional user interface
  • Solid and reliable online service
  • Outstanding customer support

  • Highly dependent on Humintell
  • Restrictive and confusing EULA
  • Slightly high pricing
  • No warranty per se
  • No stand-alone application

Humintell has the best offer on the current market, including in comparison to Dr. Paul Ekman's products. The package value of MiX 2 wins by miles and METT 1 images are free on request. There is a huge preference for Dr. Paul Ekman for his innovations and publicity brought by the TV show “Lie To Me”. Let's not lose perspective here because Dr. David Matsumoto is also a leading expert in the field, and Humintell pours more money and time in the development of their softwares and is considerably more technology oriented. Provided that you want great work from Dr. Paul Ekman, I would instead recommend to read his books, FACS manual and investigator's guide, and his research papers.

There is an eminent dependency on Humintell considering that MiX 2 is offered as a web service and it comes with prohibitive EULA. It is derailing to pay for a product that comes with no guarantee whatsoever, while we expect a world leading expert to oversee and approve every bit of the content for its accuracy and be held responsible for it. Users are prevented to keep any content they have paid for. Paying top dollars for this may make some people reconsider. I would much prefer a stand-alone application with more freedom at that price range, like many others would expect.

There are regular coupons offered, METT 1 images for free, and various other offers to increase the value of your purchase. All of them worthy on their own.

Misreading emotions is one of the chief reasons causing conflicts among human beings, let alone in macroexpressions. The training given by Humintell allows its users to develop the ability to properly read universal emotions, whether they are macro-, micro- and subtle expressions. Their entire product line deals with a subject that may change your life by increasing your ability to properly decode emotions and thus act or react accordingly in any situation.

Furthermore, even though I already have skills in non-verbal communication and reading microexpressions, I felt that the software was useful to me as well. The high quality slow motion videos allowed me to pay attention to more subtle details than I could normally do on real situation videos. Reading microexpressions in profile view was also a plus to me because it is something that we don't have that much opportunities to do since we mostly face the people we are communicating with.


Humintell is serious about quality and offers appealing products. MiX 2 is an amazing product with a life changing content and a meaningful improvement over its predecessor, and the price tag is acceptable by all means. Even though it is oriented toward newcomers, it is a good complementary training for people who already have experience reading microexpressions. MiX 2 lived up to most of my expectations and surprised me in some other aspects. The few shortcomings should not overshadow what the software will bring to your personal and professional life.

I highly recommend Humintell's Advanced Microexpression Recognition Training 2 software.


  1. Here is a quick update. According to customer support, the proficiency levels are:

    Proficient: at least 80% at a speed of at least 200msec
    Expert: at least 90% at a speed of at least 200msec
    Master: at least 95% at a speed of at least 100msec

    Now it's clear. I've had 93% at 100ms on my first run. That's 2% short of mastership! I need more training. ;)


  2. Hi. I´m about to finish studying Psicology and i´m interested in studying a Master or a degree in Microexpressions and nonverbal language. I would like to know if you could tell me where can I find information about this.
    Thank you!