Q1 has been a quiet time here at Harem Token, with most work focused on building out the infrastructure needed to develop the products on our roadmap (both personnel infrastructure and also back end architecture). To that end, now that Q2 is here upon us I wanted to take some time to go over how our Q1 goals have been met/not met and what our plans are for Q2. So the sections of this article will be; Q1 Goals recap first, Q2 goals outline second.
- V2 Head ‘n Shoulders: One of Harem’s goals for Q1 was to scale up our “Head ‘n Shoulders” product, as the V1 was very low res (256 x 256) and could take hours to load. Ultimately this task could be broken down into 3 parts; increasing resolution, decreasing loading times, and making the product available through our studio. To that end we have succeeded in the first two of these three tasks. Our resolution on HnS has been successfully bumped up to 512 x 512 and loading times have been decreased to a matter of minutes (if not seconds). A video update on this can be found here. We have delayed launching the API as we wanted to get the algorithm as advanced as possible before launching it. We are doing one final round of training for it this weekend, after which we will work on launching the official HnS algo and using it to create new source videos for our text/audio to speech options. No specific date, but you can expect the API to launch in early Q2.
- V1 Deepfake API: We have some very good news on this but don’t want the big announcement to get lost in the shuffle of things, and are still wrapping up some last minute stress testing. Expect a formal announcement on this between April 7th and 10th with the launch of the API coinciding with the announcement or coming shortly after.
- Hiring Fullstack Developer: The wonderful Peter has joined our team and has been integral in helping us build out our system and infrastructure.
- Breeding Rewards: This is now 100% finished on the back end. We have yet to roll out the API but that will be coming in Q2. You will be able to select two NFT models which you have in your wallet, breed them, and mint the resulting offspring as an NFT which can be staked. In fact, we are going to release this such that you can mint ANY two NFTs in your wallet. Want to breed a Hashmask with an Axie and mint the result as an NFT? Go for it. However in terms of staking them for use in our system, results may vary if you choose not to use 2 Harem generated girls.
- NFT Ramp Up: We exceeded our goal on this with 40 NFTs minted and 27 available for use on our text/audio to speech on our platform. The remaining 13 will be added within the next few weeks as we launch deepfake, HnS, and also our new Gen 2 NFTs.
- Randomized Waifu Generator: It turns out luckily that our wonderful farm partners at Unifty were already developing this when we announced it! They have a randomized NFT distribution service launching in April which we will be using for this once it comes out. Though FYI with the new NFT distribution schema we may reduce this number from the original 150 so that we have a sufficient number of new NFTs for our farms.
- Launch of Text/Audio to Speech and Studio UI: This was not explicitly spelled out on our Q1 roadmap, but we also want to take the time to remind everyone that in the last quarter we have developed and launched;
- A fully functional payment system.
- A point-and-click UI/UX for our studio.
- And functioning text and audio to speech systems which allow for users to write out a message, or record/upload audio and have lip movements track to their avatars.
This segment will be broken up into three sections; UI/UX, Tokenomics, AI, and KPIs.
- Having polled our telegram community we will be expanding to Binance Smart Chain. To be clear we are not migrating exclusively to BSC, but BSC will be the first non-ETH chain we expand to. So all the capabilities you as a token holder will have through our current infrastructure on ETH, you will have on BSC. This includes; staking your $HAREM with collateral to earn AMM fees, farming NFTs with your LP token, paying for the use of our studio, and receiving a discount for the use of said studio if you have $HAREM in your wallet. In Q2 we will expand all these services to a copy of $HAREM (which we are tentatively calling $bHAREM for now) which will be on Binance Smart Chain. This is in direct response to elevated gas fees on Ethereum, which we feel make using our system less pleasant for users and token holders both.
- Launching V2 Head ‘n Shoulders in 512x512 resolution. Pretty self explanatory, was detailed in the Q1 section of this article.
- The staking and unstaking process for (AI generated and/or already approved) NFTs will be made automatic. As of right now if you want to unstake your NFT you have to message the admins and deal with them directly. That’s a hassle for users, so we are automating the process so that it only takes the click of a button. This will apply to all AI generated avatar NFTs, and any approved “IRL” people who stake their faces will be white listed in so their NFT can be auto-staked or auto-unstaked through a simple point-and-click user interface.
- We will update our website to include a simple point and click interface to our farm, a Harem Token specific NFT marketplace, NFT staking and unstaking, and randomized waifu packs.
- New website design in general. It was fine for our initial launch but needs a revamp.
- Breeding system as referenced in Q1.
- Q2 avatar count goal is to have at least 100 fully usable avatars.
- One of the issues we’ve run into with big buyers is that they can’t really make any serious purchases of $HAREM without mooning the price and getting rekt on slippage as we do not have deep enough liquidity in the ETH pool on Uniswap. We want to improve this by trying to bump up liquidity in the ETH pool to at least the lesser of $500k or 150ETH pooled.*
- As previously discussed, we will be providing another alternative way to buy $HAREM through Binance Smart Chain. The working ticker for this symbol at the time of writing this is $bHAREM. This will not affect the overall token supply. For each $bHAREM we will burn one $HAREM. We will be starting out by burning 10,000 $HAREM and minting 10,000 $bHAREM**.
- *Current plans to bump up liquidity in our ETH pool are as follows.
1) OTC deals: We have worked with Lock Finance to add our ETH-HAREM UNI-V2-LP to their lockable asset list. We will use this to raise a round in which buyers will be able to send us $ETH, have a portion of it used to purchase $HAREM at a discount to market, and have the remainder of their $ETH paired with $HAREM in the Uniswap pool to provide liquidity. The liquidity will then be locked for a period of time (TBD but likely a year) with the buyer marked as the recipient of that lock. Community, models, and contributing artists will get first bite at this before we open it up to outside investors.
2) Liquidity mining. We are still working out the particulars of this but it could go one of two ways for the ETH side of things.
2a) We would use either 0xmon or Unifty’s “ERC-20 NFT Wrapper” to mint NFTs which can be later burned/exchanged for a given amount of $HAREM, and just add them to the existing farms. This is the option we are leaning towards.
2b) We would launch our own liquidity mining contract in which LPs can stake their LP through this contract instead of on our NFT farm, we have been gifted some contracts which can do this and are also talking to the devs who developed our payment system and will have some updates on what direction we are taking on this in the next few weeks.
- **Initial liquidity increases will almost certainly be funded entirely through OTC sales similar to what was discussed in the last paragraph.
- MEN. We couldn’t let Hardwood continue to go deprived, so we aim to have a working beta on a GAN to create hot sexy men by the end of this next quarter.
- As an AI company with various machine learning models for different tasks being trained we constantly are in need for more specific data. Given Harem’s focus on generative media in the adult entertainment sector the kind of labeling we need is highly specific, and tools that are commercially available have only the standard labeling capabilities. Hence we started out creating our own suite of video specific labeling tools in Q1 on which we are further going to make significant improvement in Q2 so enable faster and more accurate labeling. This is going to allow our workflow to use data generated by the deepfake API to be directly included to the training sets and through a positive feedback loop increase the accuracy of all our algorithms. This is probably meaningless to you unless you actually work in ML, but trust us it’s super important.
- Per our updated NFT/Avatar distribution schema, we are setting a goal of adding 100 new usable avatars to the studio in Q2 (not including IRL people who stake their face”).
- 5% non-occlusion based error rate (or less) on a random sampling of a dozen POV porn videos through the deepfake API. This is measured by uploading a dozen 1 minute POV videos through the API, reviewing them frame by frame, and achieving a non-occlusion based error rate of less than 5 of every 100 frames. To be clear ‘occlusion’ means an error that results from an external object or image partly or totally covering the face, or the face being partly out of frame in the video.
- Launch of Head ‘n Shoulders V1 API in our studio at 300x300 resolution. This will enable users to take simple videos of themselves from the shoulders up and create a short animated video of their chosen avatar mimicking those movements (including lip movements if they speak in the video).
- Successful beta of frame-by-frame settings adjustment AI. This can be considered one of our two ‘flagship’ product launches of Q2, and essentially boils down to having the AI able to go through the merge settings in our deepfakes on a frame-by-frame basis and optimize at least one setting around at least one error it has been trained to detect and improve on.
- Similar to the deepfake API launch, we are going to push the details of this to a later announcement so it doesn’t get lost in the shuffle. But in the next week or two we will be announcing a partnership with an anime/hentai focused NFT project in which we will be working with them to bring their waifus to laifu and make them stakeable through our platform. Stay tuned!
- Process 1,000 minutes of Deepfake orders through the API by the end of the quarter.
- 50 addresses LPing $HAREM/$bHAREM by the end of the quarter OR at least 51% of circulating token supply actively staked for LPing (not including locked tokens).