The following is a paid review of a generative AI service aggregator, called Syntx AI. I was paid to write the review, and given free credits to try out their systems, but I was not directed as to what to write, and the opinions expressed are my authentic opinions just the same.
To be honest, I was prepared to not like this service. I am a picky and opinionated internet user, and even more-so when it comes to tools that I use regularly for creative production purposes. Whether I’m working on my AI-assisted books, or other projects, I have a particular vision I’m usually chasing, and I don’t want to have tools stand in my way when I’m caught up in the rush of trying to execute on an idea that has just come to me and is screaming to be brought to life.
I also often end up in a love/hate relationship with these tools, as a heavy user as well. I tend to run into barriers quickly, with few options to resolve them or change how the tools really function to better suit the work I want to do with them.
I mostly did not hit that too much with Syntx, though. Mostly it worked how I needed and expected, as a long-time user of a myriad of generative AI tools across the board. Having many of those tools accessible in one interface was also pretty handy in the end.
One snag I did hit with them early on in my testing was with the Suno interface, where there seemed to be inadequate options to add both a style description and song lyrics. I brought this issue to the rep at the company who had reached out to me about the review. And even though they never promised to change it, they did increase the character limit on the text entry field a few days later, and I was able to include all the content I needed to prompt the Suno model. Also, perhaps this won’t last, but you’re able to download Suno audio file output directly, which I think is a privilege you now have to pay extra for on the actual Suno service itself. So that’s something to consider, if you’re into making music with AI services…
Other things I ran into were smaller things: like it seems that when you generate images in a thread, the conversations are not stateful (unless I’m doing it wrong), as they are in ChatGPT… where you can get an image output from the system, and say “change xyz” and it will just do that. I ran a test just now to make sure I’m not saying the wrong thing. But my test went like this:
- Open the Syntx image generation tab
- Leave the model on Nano Banana
- Use the prompt “generate a picture of a cow”
- Got this output:
- Then I used the prompt, “now put a hat on it.” And instead of a cow (expected behavior), I got a dog in an armchair with a hat on it:
- Then I prompted as a follow-up, “No, on the cow.”
- And I got this image of a wooden sign that literally says, “NO ON THE COW”:
All in all kind of a funny test, so seemed worth including. I suppose there must be some setting Syntx could implement to retain state between rounds for image generation? It would certainly come in handy, but I actually did not find it a deal-breaker during the work I did. Once I realized early on, that was a limitation in the tooling, I just worked around it and got results like I wanted. However, power users are going to miss that feature, unless Syntx can implement it.
But it’s things like this that are both vexing and satisfying about Syntx at the same time: because you have so many models with different options or interaction paradigms people are used to, things are bound to slip through the cracks – essential features you might have come to rely on or associated with a given service or tool in its “official” form might not always be implemented with full parity in Syntx. But based on my experience at least, if you write to the company with a legitimate use case, they will at least consider it, and maybe even implement it.
Certainly there are many elements of the Suno interface that just are not replicated in Syntx, for example. But, at the same time, you actually seem to be able access Suno through this service, and still get much of the same results for certain uses as you would if you were using Suno’s official UI directly. And since everything is credit-based (even though credits and usage limits are kind of the bane of my existence as someone creating at high volumes with AI tools), you don’t have to waste money on a bunch of subscriptions to different services, if all you need is a handful of songs here, a few seconds of video there, some text and some images over here, etc.
That said, another place I spotted a gap was in Syntx’s implementation of the Eleven Labs voice tooling. But for the short video I made using Kling via Syntx, even though I could not describe the kind of custom vocal delivery I wanted in Syntx, I was able to go over to Eleven Labs, and on a free trial do exactly what I need in one-shot without paying them anything. And in a way, this kind of bouncing back and forth is very much the name of the game for the emerging breed of AI-native creator… I think we still live in a very chaotic landscape as to tools and processes, and things are still evolving and expanding. You kind of have to exploit to the maximum potential the tools you have access to at the time and under the conditions to which you have access to them, with the understanding that the offering is likely to change in a relatively short time.
But the nice thing about Syntx is that for one subscription, you can do a great deal of your bouncing around inside one common UI, with access to many different models of different types, yielding good quality results – if you’re able to understand and work within the limitations and constraints of the implementation as you go.
Switching gears for a minute to the more administrative side: there are some things that seem to me worth noting in Syntx’s public offer/service agreement here.
- This section below seems to be granting commercial use of the generated content (though the extent to which they can even grant that is probably debatable across jurisdictions):
3.3. Through Syntx AI, the Customer has the ability to order from the Service texts, images, music, and other types of content in various neural networks, as well as to use the resulting materials for personal and/or commercial purposes, provided that all terms of this Agreement and applicable legislation are complied with.
- The company appears to be registered in the United Arab Emirates, so I’m not sure what possible complications might arise, and what “applicable legislation” might mean here. If you’re not in the UAE, and you do run into some kind of conflict with the company, or you’re in a country or a corporate situation that requires a certain category of compliance, you owe it to yourself to find out more about whether or not any of this is a deal-breaker for your use.
- Sections 6.7 through 6.10 seem to set up stringent rules around subscriptions as auto-renewing and requiring more than 24 hours before the renewal period. I did not pay for the subscription or tokens that I was granted for review purposes, so I unfortunately cannot speak to how easy or difficult it is to cancel without auto-renewal happening. Unfortunately in my experience, if a company does not have good and prompt customer service, and has not well-integrated subscription or account cancellation functions into their customer dashboards (which is a regrettably common oversight), things like this can become headaches for users. On that note, it is also worth reviewing their refund terms under 7.1 to 7.7 in the public offer document linked above. The wording there makes it sound like they do not like giving out refunds…
There’s possibly more to be discussed in the Public Offer, but I wanted to pivot to the Privacy Policy, since it impacts some product UI functions in not entirely clear ways. Among other things, it lists the following as being stored by Syntx:
2.2.2.1. History of messages and commands sent by the User to Syntx AI
However, in the UI when you are entering a prompt, there is an :unlock: icon that if you click on it changes to :locked: and says “We’ll save your prompt” in a drop-down notification. And if you’re in the :locked: icon state, when you click again to unlock it, the drop-down says “Your prompt will disappear after sending.”
So my expectation using that is that the unlocked state becomes something like a temporary message in ChatGPT or elsewhere, where the message contents get erased some time after use. But when I go back in the list of prior chat sessions, I can still see my prior prompts and results. So I’m not really sure what is happening: are they being stored or not? Does the actual product functionality quite reflect what the Privacy Policy says? I would say it deserves closer attention from their Legal & Product teams to make sure it is all implemented in a way that is compliant with their objectives and applicable law. If you as an end user have elevated compliance needs, you would also do well to perform greater due diligence to make sure the company’s retention and other practices align with your needs. (It also says, for example, that your interaction records with the service are “stored for 12 months from the User’s last interaction with the service.”)
I don’t want to go bananas over their policies, because I got a lot of utility out of the product for my pretty niche use cases. But there are certain things that I don’t like all that much as an end user, like having marketing cookies that track my interests for personalized advertising:
8.2.1.3. Marketing cookies that may be used to collect information about your interests and preferences in order to personalize advertising or content presented through Syntx AI.
That’s not too exciting. If I’m paying, I ought to be able to turn off and opt-out of things like that.
Depending on what kind of user you are, the above administrative details might or might not get in the way of your ultimate enjoyment of the service and its features. They are red flags for me, but not deal-breakers.
I looked around a bit to see if there are other red flags based on public reviews by other customers. There are a few negative ones on Trustpilot against the company. Of the two and one-star reviews left by hopefully real customers, the complaints range from things that are a bit inscrutable (about HR practices in hiring, and the creator rewards program), to some missed expectations about what subscription actually entails, difficulties with the refund process, and claims that the service is slow or unreliable.
I did have a couple of video generations fail outright at certain points. I will say that. A couple of them were because of inadvertently triggering content filters, but at least one of them was unexplained, and just hung and never resolved with a meaningful error message. But that’s also kind of the life of someone using AI to produce things like this. Sometimes the tools just don’t work and we don’t know why and there’s nothing you can do about it.
As to speed overall, I wouldn’t say generations are extremely fast in general, but I didn’t find them particularly slow either considering they are running over APIs, and not themselves natively hosting these.
In any event, the company did at least seem to respond to the Trustpilot reviews, for whatever that is worth. Poking around with the help of AI search, ChatGPT found me a couple other sites with Russian-language reviews of Syntx, including this Russian Apple App Store page, where the auto-translated reviews for the mobile version are not great. I should specify I’ve only used the web/desktop version, not mobile myself, so I can’t speak to that product experience.
Russian-language results on sites.reviews also list the following points which appear to have been auto-summarized out of the published reviews under “What I don’t like”:
- Issues with subscription cancellation and auto-renewal
- Technical failures and service freezes
- Slow support and lack of assistance
But, under the “What I do like” column, we see basically the same points that I agree with about the service (given that I have not had to deal with their billing systems):
- A wide selection of neural networks for various tasks
- Intuitive interface
- Affordable prices and free tokens for testing
- Convenience of working with one subscription
Possibly helpful tangent: I did notice the other day that Muse AI uses something called Stripe Link, which is a way to have single-use authorizations of temporary virtual cards that get used by AI agents for a specific purpose only. Perhaps if the ability to easily shut down unwanted recurring payments is an issue for you (and let’s face it, this is universally annoying), going down a similar route might be desirable. I know when I’m trying out new services, I will often associate a Google Account with them, because then later I can just go into my Google dashboard and sever the account link there, rather than try to deal with varied account deletion processes across different services. So perhaps a single use or disposable card number could do approximately the same thing for enabling some greater level of control over recurring billing in a system like Syntx’s. But it’s not something I’ve tested at all on my end, just tying it in here because its inclusion in Muse proves that it is a “thing” on many people’s radar.
All that said, it seems like many public reviews suggest people are not happy with aspects of automated billing and refunds, and the company would do well to look at whether or not they currently have the best policies and workflows in place around this. However, all that said, according to their FAQ on the bottom of the homepage, it says to disable auto-renew, go to your Profile > Cancel auto-renew. I can’t verify that myself, as I am not using a paid account for testing. Mine instead says Elite membership, subscription end date, with two buttons: [Renew subscription] and [Change plan].
Let’s see what else [checks notes]…
Oh, one thing I want to point out, the workflow example I posted the other day about how to make video, I would not probably have easily been able to come to that formula were it not for Syntx offering me all the applicable tools right in one place.
So I think there’s also something to be said for unexpected opportunities that arise out of having the tools all together in proximity like this.
I experimented a bit with the “Agents” section of the Syntx service, and will be honest that I could not quite get it all to work as expected. First, I tried making a group of individual agents who had different roles for writing, wrecking, and editing book manuscripts, but that didn’t go anywhere at all, and was not obvious to figure out.
Many days later, after realizing I could ask Codex to produce a completed video out of the individual elements I uploaded to it (video clips, audio voice-over, script) and a general instruction of what to do, I tried reproducing the same thing in the Syntx agent interface. I got significantly farther than I did on my first pass, and realized I could use the agent to invoke other tools linked in the workspace. Here’s a screenshot of what is offered (GPT-5 is the highest available listed model as of this writing in the drop-down)
I was able to direct it to invoke the other tools to some extent. My system instructions put into the bot’s setup were:
This agent orchestrates other AI services to generate a short video script based on a premise inputted by the user. The agent then generates a text script summary using GPT-6 or highest available creative text generation service.
The generated script is broken up into blocks intended to be between 6 and 10 seconds of video each, with an accompanying prompt to generate the video in a high quality generation service such as Kling.
The agent oversees then execution of still frames based on each video block that visually depict the contents of the script in a visually interesting and appropriate style and manner, and ensures that taken together the various generated video segments fit together in the same world as identified in the script and initial premise.
Agent then detects if there are narration needs or other unaccounted for audio artifact needs (such as music of sound effects), and accesses Eleven Labs for speech synthesis and/or Suno for music generations.
Agent then prepares for export all generated materials, including: premise, summary, script, visual and video prompts and their image and video file results, audio prompts and their audio file results.
Well, reading that again, I can already see some issues with the prompt (in terms especially of how I chain together the script > keyframe stills > video segment output), but I did a lot of coaching and hand-holding in the thread, and was never able to get close to that whole package as a “smoke test.” I was able to get bits and pieces of it: I got some still frames, some text descriptions of what happens in a block, some video segments, but compared to just running the same overall workflow myself manually, it was not a success. I didn’t end up with a packet of files to download for Codex to assemble. And I gave up before I ever tried to get my Syntx agent to try and assemble it all into a finished video (a capacity I’m not certain it has, anyway).
After having come over from experimenting heavily with Codex where I’m having increasing success (through trial and error and refinement) getting things to “just work,” the Syntx agent experience left a lot to be desired. Should I expect an offering like this to be as full-featured as Codex *AND* include all these other models for the same price? On the one hand, maybe not. But on the other: Yes? Like I said, I’m a picky end user. Why can’t I have the best of everything at a great price? As models develop and use patterns mature over time, I think that this will be the main competition in the market: every service will be able to offer everything. Which then, will you choose, and why?
For now, of course, Syntx still offers an aggregation of other models and services that you can’t really get from ChatGPT – though they seem to have plenty of competitors in the multi-tool AI aggregator space.
I still find the pricing a bit weird, since you have to have an active membership at whatever level (which might include unlimited use of certain included models), and then buy tokens on top of that… I guess for the things that require tokens. According to their FAQ, if you have tokens after your plan expires, they don’t go away, but you need an active plan to use them… One other thing I found confusing: when you pay for subscriptions, it seems to be billed in USD. But if you later buy a la carte credits, it wants you to pay in Euro, even if you select USA as region… This seems like a bug, and also makes pricing relative to credits even more opaque by mixing currencies.
Whether Syntx is the best aggregator tool like this out there right now, I cannot honestly say, as I have not tested many of the other tools competing in this space right now. I do think the concept is cool, and the market seems to still be developing. I think whichever one you end up choosing, there are likely going to be hiccups, though they may be different than the ones you’d see here.
If you think you want to try out the service, the company gave me a promo code (which I do not earn revenue from) which will give you 15% off any subscription plan, according to the company. Promo code to use at checkout: LOSTBOOKS.
If you have good or bad or middling experiences with the service, please send them along to me and I will do a follow-up post.
Leave a Reply
You must be logged in to post a comment.