TaChiKu: Deepfake NFTs

[IMPACT] Please describe your proposed solution.

Starting point:

This project started in Fund 7 as a simple concept: To offer any user of Web3 Apps / Metaverse, or even web2 Apps (twitter, discord…) a simple way to obtain a fun picture profile (PP) of themselves matching the context. If I play a game of zombies, I d like a PP of me as a zombie, instead of a generic zombie. As the digital world grows and end-users have dozens (maybe even hundreds) of online identities, they will want dozens (if not hundreds) of PP to match moods, stories, and situations.

A first iteration was developed in the past 5 months (see TaChiKu.com), and acts as a very good proof of concept. It works, it's simple to use, it's fun, and it is also deeply addictive (just try it out, you will see what we mean). At time of writing (end June 2022), the minting is not yet integrated, but there is still 6 weeks until final delivery of this "first iteration".

The promise made in November 2021 was

> "A web3 API for generating an ADA compatible NFT where the picture is generated by an AI. Input an image of you, a few words for context (zombie, punk, etc..), your ADA public address, accept to pay a few ADA for minting, and your PP appears at the other end of the API call."

A morphing of an uploaded image

https://tachiku.com/app/view/?uid_image=84469401340

TaChiKu.com does this (*minting coming by mid August latest) although we went way beyond "just an API" and instead provided a full web app for an easy end-user experience. For devs, the API integration here: https://tachiku.com/doc/

However, the current solution has both some "limitations", and some higher value "potential". This is what we would like to address in this proposal.

The current limitations:

1/ Difficult to setup

Setting up an AI (from an open source project) is time consuming and can be very difficult, as it will depend on OS (Windows, Mac, Linux), GPU (Nvidia, others, drivers), and compatibility of open libraries in the various combinations of OSes/GPUs.

In the past 5 months, we have delivered a morphing AI which works well for either Windows 10 + Nvidia RTX3090 or Linux Ubuntu 20.04 + Nvidia RTX3090. We would need to deliver more AIs (including the delivery of the Deepfake AI, but also potentially other image manipulation AIs) for more GPUs, and for most latest OS releases.

Delivering this part is a necessary pre-condition for the next one (2/).

2/ Slow to warm-up

Most AIs run from Python programs, and they take a very long time to warm up, therefore this is not a workable solution at scale. Recently, we have transformed the Morphing AI from a Python program to a local server. There needs to be a systematic approach with open APIs for moving Python programs to server based AIs, so that the integration is smooth.

Delivering this part is a necessary pre-condition for the next one (3/).

3/ In need of virtualization

The robustness, redundancy, and capacity under load of the entire TaChiKu solution requires several concurrent AIs running in parallel. The only good way to do this is via virtualization of those AIs.

Anyone wanting to lend GPU power should not have to worry about the 2 problems above. They should basically just drag and drop a docker image, start it, and be accepted as part of the TaChiKu AI network within seconds.

Unfortunately, the containerization of AIs is still machine dependent (in fact it is GPU dependent + python libraries may require adaptation / recompile on yet unsupported combinations of GPUs, and also Docker can make use of the physical machine GPU power only on linux…). Although some attempts at dockerization were made in the past 3 months, none were fully successful yet, and more times needs to be spent here.

This part is critical to move to the more interesting "value potential" section.

To understand more, watch this video explaining the architecture journey from initial Proof of Concept to where we need to go for a full virtualization of AIs:

<https://www.youtube.com/embed/4T7wFS0xONY>Note that because of the lack of an easy way to share and setup a working AI (via container), the current project is limited in the sense that the AI only runs on our physical machine. The TaChiKu app itself is hosted on AWS and available 24/7, but the morphing AI program is physically hosted on our machine. When the machine is Off… the Morphing AI is Off, which of course is not a very acceptable scaling proposal. It is only good for the current stage of a final Proof of Concept.

The value potential:

When the AIs are containerized and run on their own server, able to communicate back and forth with the main TaChiKu AI server, it then allows for an ecosystem of fully decentralized AIs. With this in mind, the potential of TaChiKu could go way beyond its initial scope.

The development effort submitted in this proposal would deliver a list of Docker images ready to be instantiated in pre-defined environments (AI + GPU), for example:

"Morphing - RTX3090 GPU - tested on Linux Ubuntu"
"Morphing & Deepfake - RTX3060 GPU - tested on Linux Ubuntu "

Those would be freely accessible to anyone. You could download the image, run it on your machine (one command line to execute) and it would be already pre-configured to automatically connect to the main TaChiKu AI server, which would automatically integrate your machine's GPU power into the whole decentralized TaChiKu AI ecosystem. Morphing and deepfakes executed on the TaChiKu app would then be sent to an available dockerized AI for processing, potentially yours if it is running and available at the time of the request.

You may ask: why would I want to participate into lending my GPU power? Well, we can imagine an eco-system where the participating user (lending his GPU) could be rewarded in coins taken from the end-user's fee for morphing and minting NFTs.

Note: this last part of rewarding GPU owners is not part of this proposal, but this current proposal would put in place all the bricks for this to happen thereafter.

With this proposal, TaChiKu can become a fully decentralised AI ecosystem for minting AI-generated NFTs on Cardano. Think of all the possibilities: morphing, deepfakes, generative art, a long list of whatever the combination of end-user's ideas, GPU power, and AI capabilities can deliver.

> In short, we can give to anyone on the internet, within a few clicks (and a few ADAs), unlimited access to complex pre-built AIs alongside with the GPU power to run them, and the instant minting of the result on the Cardano Blockchain, as well as automatic posting on social media.

And later, we will be able to give all lenders of their GPU power a reward / stake in the ecosystem.

What we will deliver:

1/ a robust virtualized network of curated AIs (morphing, deepfake) running on nodes hosted by participating community members.

2/ the ability for end-users to generate morphed images and deepfake videos from a simple image upload.

a 15sec example deepfake made post morphing (see picture above)

<https://twitter.com/TaChiKuAi/status/1541508014959726592>

3/ the ability for end-users to mint those morphed images and deepfake videos as NFTs on the Cardano blockchain, and automatically post them to relevant social media platforms.

4/ the ability for end-users to trade AI-generated artworks as native Cardano NFTs on a dedicated marketplace with like-minded users.

Benefits:

This project will benefits the wider Cardano community in two ways:

1/ As TaChiKu continue to deliver simplicity of use of AI capabilities and access to GPU power, it can become the global go-to platform for AI-generated NFTs. It would only take a few influencers to kick start a frenzy. The trend of avatar personalization, memes, and short but impactful videos is only going to grow, and TaChiKu has all the building blocks to be very well placed and catch this unstoppable wave. In this sense, an obvious benefit for the Cardano community is the exposure to gain large communities of end-users (and the associated fees), as well as being positioned in this growing niche market of AI-generated artwork.

2/ If the project attracts a large volume of end-users, it will then attract "lenders of GPU power" since they will have a financial incentive to be part of the network. Those participating lenders will add to the stickiness of the solution and to the overall value of the ecosystem (think of a future Utility coin) which will also greatly benefit the Cardano ecosystem.

[IMPACT] Please describe how your proposed solution will address the Challenge that you have submitted it in.

From an end-user point of view, TaChiKu can become "The Marketplace for AI generated Art on Cardano". Although its DApp will ultimately be fairly simple and common-place (mint, buy, sell, and possibly auction), its appeal will be in its unique access to AI "magic" to create surprising images and videos within a few seconds. All mints and subsequent transactions (resale) made by end-users will benefit the Cardano ecosystem in fees.

TaChiKu will be positionned as a unique decentralized AI solution for the Cardano ecosystem, to benefit from the growing market of NFTs, avatars, AI driven artwork, generative Art, and more. The niche of AI driven art is still in its infancy, and the immediate attraction of deepfakes and personalizable video memes could be an instant hit, attracting a few influencers to mint themselves on the Cardano network, and opening viral snowball effect.

TaChiKu will directly plug into social media, and any take up will be massively amplified, thereby benefitting the Cardano community. Have a look right now to see how it works: try morphing your picture on TaChiku.com, make it publicly available, and you will see it on twitter within seconds - follow @tachikuai

Finally, TaChiKu could later on deliver its own Utility token on the Cardano blockchain, increasing TaChiKu's dependency on Cardano, and allowing Cardano to benefit greatly from any success made by TaChiKu.

[IMPACT] What are the main risks that could prevent you from delivering the project successfully and please explain how you will mitigate each risk?

Here is the list of the identified risks for the next 6-12 months:

Technology risks:

The risk on the backend and web app are mostly contained, since the experience of delivering the first iteration of TaChiKu have been positive. The biggest risk here is therefore not so much of technology, but of time (Eric's time). At this stage, it is still planned that this part of the project will be delivered by Eric (to lower any risk of ramping up another person on this core part of the project). In order to mitigate the risk, the project timelines have been expanded to 9 months for the final delivery. There will be many intermediate milestones, but at least, the 9 months timeline will act as a buffer to mitigate any risk on availability of Eric's time.

The risk on the Plutus code dev is also contained in its own box. There is no question that such code can be delivered, since NFT minting is not a new feature on Cardano. It is just about finding the right person. In the unfortunate case that this part could not be delivered (which seems very unlikely), we would continue to work on the third party solution (nft-maker) for a while. So it does not break the whole project if not delivered.

The risk on delivering a fully decentralized AI network seems the most high profile technology risk of this project. It seems easy on paper (dockerise the AI and use it) but it comes with many constraints of integration, that it was not possible to deliver within a few days so far. This risk is best mitigated by finding the right person for this job. At the time of first submission of this proposal, this person is not yet identified, but it seems very reasonable that we will find a Python developer with relevant docker experience to help us for a month or two, and deliver the final dockerised versions of morphing and deepfake AIs.

Commercial risks:

Then, there is the risk of the project not taking up. So far, experience shows that end-users love it when they see the result of image morphing. By going one step further and delivering deepfake videos, we hope to go beyond the initial wow factor and attain long-term stickiness. By integrating with most social platforms (twitter, facebook, tiktok, linkedin), we indeed hope to get to a point of a virtuous cycle of free viral marketing. We have allocated a marketing budget to mitigate this risk and turn this project into a success with a large end-user adoption and hopefully lots of goodwill viral marketing. The recruit of influencers will likely be key (we are currently discussing with some artists with a very large fan base)

Financial risks:

Since this project has been initially self funded and helped by the Cardano community (fund 7), we do not depend of VC money. This project is not at risk financially, since the outgoings are so low that the grants and the self funding from Eric are enough to keep it going for as long as we can see. If this grant is not given, then the project will continue, but at a very slower pace, and at the cost of some "must have" being pushed back and/or discarded until funding is in place. Note also that beyond the funding from Catalyst, the access to the Cardano community also gives us a strong momentum and advantage.

Single chain risk:

It could be argued that there is also a risk associated with keeping the project only on the Cardano blockchain, and that opening it to other chains (mint on Solana, on Eth, on Cosmos, etc…) could be a good move, and make the project so much more visible. In the last few months, when the integration of Cardano NFT minting (plutus DApp) was becoming "much more painful" than initially anticipated, we could have reverted to other solutions and decide to mint on another chain with a potentially "easier" implementation. But we did NOT do this. Instead, we continued to search for a solution that would work on Cardano (and we found it).

We want to express openly and very clearly where our loyalty goes. With past and (hopefully) upcoming additional funding coming from Cardano and supported by its community, this project for delivering Morphed and Deepfake NFTs via AI will stay on Cardano and not go anywhere else.

> Loyalty is an asset that we will not exchange for a quick win elsewhere. We have embraced this community, and we are looking forward to a long term win-win relationship together.

[FEASIBILITY] Please provide a detailed plan, including timeline and key milestones for delivering your proposal.

Alternative commitments

Before providing the detailed plan, please note that the main person behind this project (Eric) got the privilege to receive two project funding in the past, and is pitching for another project in Fund 9:

Fund 7 - https://cardano.ideascale.com/c/idea/383844 - as described in many sections of this application, the current proposal is a continuation of this TaChiKu project initially granted in Fund 7, which will be delivered within a few weeks. Since early 2022, it was planned for a delivery on August 15, 2022. You can also check progress by launching https://tachiku.com, which has been live since around early May 2022.
Fund 8 - https://cardano.ideascale.com/c/idea/400022 - this project is just starting now (June 2022), and I have given myself 8 months to deliver (final delivery planned on 30th Jan 2023), therefore I am not working on it full time (only around 2 days per week), so I have still a lot of bandwidth remaining.
Fund 9 - https://cardano.ideascale.com/c/idea/418526 - I am submitting another project in Fund 9, which is planned for circa 460 hours of my (Eric) time over 7 months (ie around 15 hours per week, or 2 "small" days per week). This project has very minimal risks on delivery since it is mainly the re-writing of a private web2 project, that I developed 2 years ago, towards an open source app integrating with Cardano.

Between project on Fund 8 (idea/400022 ) and the other pitch for fund 9 (idea/418526) if it gets funded, it would take around around 32 hours per week of Eric's time, until around Jan 2023. Therefore, there is still bandwidth for at least another 20 hours per week.

In conclusion:

> (i) the other projects taking some of Eric's time are managed with long deadlines (no aggressive commitments) and low technical risks (mostly rewriting in web3 what was done in web2) ;

> (ii)the timelines of this proposal are set to accommodate generously all concurrent activities, so that each project is delivered as per its respective expectations (as was done with the Fund 7 TaChiKu project, delivered on time and on budget).

Project timeline

As soon as this proposal gets funded, we will start delivering according to the following schedule:

Within the first 8 weeks: delivery of the virtualization of AIs (Python / Docker)
Within the first 12 weeks: delivery of the Deepfake virtual AI
Within the first 20 weeks: delivery of the NFT DApp (mint, buy, sell)
The other parts of the project which do not depend on external experts will be paced with a workload estimation of around 20h per week. As detailed in the section below, Eric's dev workload is estimated at 85 days (or 680 hours), which represents around 34 weeks at 20h/week, or a little less than 8 months of work. Setting up the project duration over 9 months gives a good enough cushion to accommodate for risk.

Therefore the final delivery of the project is set for end June 2023, although there will be many intermediate milestones, all delivered "in production" (i.e. accessible from the TaChiKu.com prod website). There is an alternative "plan B" to accelerate this project, but it would require extra private funding not yet secured, so although we may be able to get a positive surprise boost, we keep our commitments aligned with the slower pace timeline.

In the past 3 months, the TaChiKu Web App (including all backend and AIs running underneath it) has been updated at least one time per week (frequent agile release cycle). This continuous delivery of production-ready releases will carry on at the same pace for the foreseable future.

Backlog of dev tasks to deliver:

(wd= working day of 8 hours)

[31 days] Infra (Python / Docker)

This part is a critical work to allow the delivery of decentralised AIs. Most backlog items in this section will be subcontracted to an expert in Python / Docker. It is mostly independent from other tasks.

[15 days] Python progs on server

[3 wd] API integration with TaChiKu AI
[2 wd] Full doc for Python server integration with TaChiKu AI
[2 wd] Make Docker servers connect to TaChiKu at launch
[3 wd] AI server metrics APIs
[5 wd] AI server monitor

[6 days] Deepfake integration

[6 wd] Setting up local config for Deepfake AI

[10 days] AI containerization

[4 wd] Docker for Morphing + RTX (up to 3090)
[6 wd] Docker for Deepfake + RTX (up to 3090)

[35 days] Backend

This part is necessary to make use of the Dockerized AIs. There is a need for a large refactor of the TaChiKu AI backend (move it from a localhost to the production server alongside TaChiKu main backend)

[25 days] AI Backend

[6 wd] Integrate TachiKu AI into TaChiKu backend
[6 wd] Start AI from pool (rank, choose, remediate, revoke…)
[3 wd] API integration Deepfake / TaChiKu AI
[4 wd] API doc
[6 wd] Test scripts

[10 days] TaChiKu backend

[2 wd] Minting
[4 wd] Security
[4 wd] API doc

[26 days] onChain progs

During the initial project, there was an attempt to build our own Plutus DApp for minting NFTs, but it became clear that this was a big job requiring expertise in Plutus, so the NFT minting was integrated with a third party (nft-maker). It would be good to have our own DApp instead, hence the need for this work to be delivered by a Plutus expert. It is mostly independent from other tasks.

[26 days] Own NFT plutus Dapp

[10 wd] NFT Dapp fpr minting NFT
[8 wd] NFT Dapps for resale of NFT (1 for sale, 1 for buy)
[5 wd] Integrate Dapps with TaChiKu
[2 wd] Test scripts
[1 wd] Doc

[49 days] Web App

The web app will need to evolve to reflect the changes of the 3 main projects above.

[13 days] Web App - NFT

[4 wd] Improve UI of NFT Metadata
[3 wd] Buy UI
[3 wd] Sell UI
[3 wd] Mint UI

[13 days] Web App - NFT Gallery / Marketplace

[5 wd] List All NFT available for mint – marketplace
[5 wd] List All NFT available for buy – marketplace
[3 wd] List NFT from user – owned / for sale / starred

[7 days] Web App – misc

[3 wd] UI for deepfakes
[2 wd] Sorting of lists
[2 wd] Filtering of lists (incl metadata)

[11 days] Web App – integrations

[3 wd] Facebook integration
[3 wd] LinkedIn integration
[2 wd] Full twitter integration
[3 wd] TikTok integration

[5 days] Web App – Security

[2 wd] investigation
[3 wd] User account basic features (incl GDPR related)

[21 days] Project Mgt + risk allowance

The outsourced project will require a little bit of extra project management and integration, and the risk on the whole dev project (all dev tasks) has been mitigated by a 12.5% allowance.

[15 wd] Risk mitigation on dev @12.5% overall
[6 wd] Project Mgt of outsourced work @10%

Summary of dev related activities

The overall dev budget split is reflected in the graph below, with more than 50% allocated to the top 3 sub projects (dockerization, Backend updates, onChain plutus DApps). Then one third of the activity is placed on further dev of the web app, which is the UI/UX view of those layers below (Backed, AIs, Plutus DApps).

Image File

Marketing budget

As the delivery of a fully decentralised image morphing and deepfake video making solution gets closer to being achieved on the Cardano blockchain, we would like to include a discretionary marketing bugdet to spend on promoting the app. This would include mostly:

improving the website and making/integrating a few "viral" videos
paying for GPU usage and/or minting of the first circa 500 onboarded users to build a first solid community
offsetting partially the cost of an employee recruited specifically for marketing purposes and for animating a community of users

[FEASIBILITY] Please provide a detailed budget breakdown.

A budget of $95,350 is proposed as follows:

[$65,800] Development work

[$36,400] Docker, Python, Plutus (by experts): 52 days (at $700 per day)
[$14,700] TaChiKu backend dev: 21 days (at $700 per day)
[$34,300] Web App dev: 49 days (at $700 per day)
[-$19,600] discounting 40% of Eric's dev time value for Backend and Web App dev

[$6,650] Doc and setup

[$2,800] test / setup activities: 8 days (at $350 per day)
[$3,850] Documentation (APIs mostly): 11 days (at $350 per day)

[$14,100] Project management and risk mitigation

[$3,600] Project management of outsourced dev : 6 days (at $600 per day)
[$10,500] Risk mitigation on all dev tasks: 15 days (at $700 per day)

[$8,800] Marketing

[$1,000] pay for GPU and minting cost of first 500 users onboarding
[$1,600] a couple of viral videos
[$6,200] 50% of employee's cost to animate community and various marketing (first 6 months only)

Note that the $10,500 risk mitigation cost also act for us as a buffer for the (at least recent) volatility of the ADA coin, since we will have to pay expert contractors and employee in Fiat, not in ADA, and we may not be lucky in our timing for exchanging ADA against USD.

[FEASIBILITY] Please provide details of the people who will work on the project.

Eric Duneau: <https://www.linkedin.com/in/ericduneau/> - Most coding of the backend and web app will be developed by Eric, in continuation of what was delivered so far in prod (see TaChiKu.com). The project management of the outsourced parts is also done by Eric. The workload is around 86 days, which represents 8 months at 20 hours per week. The elapse duration of this project is set to 9 months, to leave ample time and mitigate any risk.

Dmitry Shibaev: Dmitry is a member of the Catalyst community (@dynamicstrategiesio) and has gained authority in the community in this domain via his open source cardano-wallet-connector (<https://github.com/dynamicstrategies/cardano-wallet-connector>). We have agreed to collaborate with Dmitry to deliver our Plutus app. Dmitry is building a similar NFT minting functionality

in his proposals (Cardano Beam - GPS based Assets (ideascale.com)) and we will gain synergies by working together.

Unidentified yet (Python / Docker): the person for building the Dockerization of Python AI programs is not yet identified, but will be shortly. The task is very much independent from others in the project, so it can be managed quite separately from other backlog items (although it will have to plug into the backend at some point). There is good evidence that this kind of profile can be found relatively easily, to be continued soon…

Marketing: the team still needs to be built. Most of the marketing budget is for supporting the recruitment of a marketing person, which will therefore be in position after this proposal is voted on.

[FEASIBILITY] If you are funded, will you return to Catalyst in a later round for further funding? Please explain why / why not.

There is a chance to come back for one last round, although there is no certainty at this point.

It was not the intention when the proof of concept was kick-started via the grant in Fund 7, but we have since learned a lot doing this project, and it has opened new very interesting avenues. Here is the summary of the lessons gained from this initial project, and what it opens up for the future:

Stage 0: Proof of Interest

The submission of the initial proposal to Fund 7 was made very late in the process and with absolutely zero communication about it. The project got funded on its own merit. Getting the funding despite zero communication signaled that there is a tremendous interest for such project in the Cardano community.

Image File

Einstein by a Seurat AI

Stage 1: Proof of Concept

The funding of $22,500 from Fund 7 will be consumed by mid August (before this new Fund 9 starts). Thanks to this funding, we researched and implemented solutions way beyond what was initially pitched back in November 2021. By mid August 2022, TaChiKu will have become a very appealing and easy-to-consume web3 app (web app + AI bots + web3 integration on Cardano), but it will still be restricted in terms of AI (capabilities and inter-operability). Beyond the initial concept of morphed pictures and NFT minting, the project clearly opened the door to a higher value proposition: delivering a fully decentralized AI ecosystem.

Image File

Magician Einstein by AI

Stage 2: Proof of Scale

This is where we are now. If we have the privilege to get granted from Fund 9, we will use the funds to improve the capability and inter-operability of the AI, open the system (open APIs between open source AI and TaChiKu, Docker bots…), and make the AI bots fully decentralized. At the same time, we will start to make the project known in the wider community (we plan to recruit a marketing person for this). At this stage, this project will naturally scale way beyond its initial scope.

Image File

AI <s>breeds</s> morphs cats with other animals

Stage 3: Proof of Value

If/when the project scales (say within 6-9 months of having received Fund 9), it means that we have a fully decentralized AI. Therefore, it means that users participate into the eco-system by lending their GPU power. Therefore it also means that they should expect compensation for this.

At this stage, it could be argued that the underlying architecture which was build to deliver the TaChiKu project could be easily repurposed as a strong decentralized AI ecosystem, with its own utility token. Therefore, it opens the possibility for one last pitch: "the creation of a utility token, running on the Cardano Blockchain, whose main purpose will be to reward AI nodes utilizing GPU power of participating users".

In truth, it would not so much be a direct continuation of TaChiKu, but would become a utility project with its own value, which other projects, such as TaChiKu, could leverage for their need of AI apps and compute.

> This utility token could go way beyond just morphing, deepfakes, and NFT minting, as it could drive an economy of shared computational power for consuming decentralized AI, ultimately whatever the purpose of such AI.

Image File

AI draws inspiration from a Quartet to transform an Emoji

Should we pitch for a new AI Utility Token project in around 9-12 months, we will apply the same restraint as already applied in the two proposals (fund7 pitch was made at a highly discounted dev cost for less than $25 per hour, whereas Fund 9 pitch discounts Eric's cost by 40%) and prove value for money beyond doubt (same motto: under promise, over deliver).

[AUDITABILITY] Please describe what you will measure to track your project's progress, and how will you measure these?

We will have public (accessible to anyone), personal (accessible by a TaChiKu usser), catalyst (accessible to anyone in catalyst), and private (accessible by TaChiKu team only) metrics.

[public] Power of the network

The TaChiKu website will integrate a page showing the number of Docker AI nodes running at any one time, to include:

how long the node has been online for
what GPU power it can deliver
how fast it processed the last few requests
how many requests processed in the last period
how big is the node's backlog to be processed
where the node can be reached
how many AI programs it can serve

[personal] Stability / availability

The Web App website will integrate availability flags and monitoring of activity (the web app already integrates a flag - red / green - indicating the availability of the "currently single node" AI) :

is morphing available?
how long to expect for a morphing to execute
same with deepfakes (available? how long?)
later on, we will likely integrate a "cost preview" (how much it costs to morph…)
availability trace of the network in a past set period (last day, last week, last month)

[personal] Social engagement KPIs

The twitting of morphed images is already done, but we will provide monitoring metrics of social activity (Twitter, Facebook, TikTok, Linkedin) accessible by anyone from the web app.

how many tweets, FB posts, LinkedIn posts, TikTok shares…
gamification of "clans" (zombie clan vs ghost clan) => in which clan do you belong?
number of likes / dislikes on images/videos made public

[private] User engagement KPIs

This weekly monitoring will be done for our benefit but not shared with end users:

monitoring of TaChiKu usage (number of website access vs app access vs actual morphing executed)
number of downloads vs number of mints
number of images generated per user (averages, median, …)
total number of mints per period
auto-detect frequent users, influencers

[catalyst] monthly report to Cardano

We report monthly to the Cardano / catalyst community via a video report, which generally includes:

our last month activity (what was delivered)
what we will focus on during the next month (what we will deliver next)
our challenges and opportunities

[AUDITABILITY] What does success for this project look like?

Technology: the TaChiKu web app is released and stable. It is working on top of Cardano and an decentralized networks of AIs, which are powering its capabilities to generated morphed images and deepfake videos.

Value for user: those morphed images and deepfak'ed videos can be minted on Cardano within a few seconds of first image upload. There is optionality for the user to mint for himself, or to mint for resale (AI art).

Value for nodes: those providing the Dockerized AI nodes see that there is good uptake on GPU usage and minting by end-users, making them want to take part and benefit from the eco-system. They accept to share they GPU power against value, which will kick start the next and (so far) ultimate phase of this project: to deliver a utility coin and share the value with participating nodes.

onboarding nodes: the AI nodes are not dependent on one machine only, and the system has built redundancy (for a start, with at least 5 permanent active nodes for the past 30 days)

Influencers: a few influencers "play the game" and upload deepfake videos of themselves as Zombie, Joker, or whatever else, which go viral. A growing user base follows, with at least a few hundreds of users every day (circa 1,000 uploaded images per day).

Events: we have a few event-driven peaks of 50,000 uploads per day, such as for halloween for which every one seems to want their face morphed as a Zombie and minted on the Cardano Blockchain to be part of the same clan as their influencer.

Memes: a few memes made by some influencers are delivered as a 5sec video into the TaChiKu ecosystem, and they become the standard for making anybody's own deepfakes. Memes become AI driven and are now deeply contextualized/ personalized. The TaChiKu network is having trouble to cope with peak loads. The journey towards becoming a Global piece of Web3 infrastructure begins…

[AUDITABILITY] Please provide information on whether this proposal is a continuation of a previously funded project in Catalyst or an entirely new one.

Yes - This is a continuation of a funded project from fund 7 (project 700184). See https://cardano.ideascale.com/c/idea/383844

I got a grant of $22,500 to kick start this project. I worked on the project for around 5 months so far (1 more month to go as of end June 2022). The technology part of this funded project is on track (delivered lots more than what was planned in terms of technology) but I re-routed most of the tiny ($2,000) marketing budget to fund extra development and infra setup costs. Therefore at this stage, the technology is here and promising, beyond anything that was initially promised, but the user base is not yet established.

Now that I have a very strong Proof of Concept available to demo on tachiku.com, I am seeking the opportunity to push this project closer to large scale production stage through a robust decentralized AI network and a heavier marketing budget.

Sustainable Development Goals (SDG) Rating

SDG goals:

Goal 9. Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation

SDG subgoals:

9.b Support domestic technology development, research and innovation in developing countries, including by ensuring a conducive policy environment for, inter alia, industrial diversification and value addition to commodities

Key Performance Indicator (KPI):

9.c.1 Proportion of population covered by a mobile network, by technology

#proposertoolsdg

bookmarked!