Yodayo combines anime image generation, Tavern character roleplay chat and a creator community on one credit-based platform, with a Model Hub of user-uploaded checkpoints and LoRA training for consistent characters.
Used this tool? Rate it
Used this tool? Rate it
Yodayo is an anime-focused creative platform that puts three things behind one login: an AI image generator tuned for anime styles, a character roleplay chat system called Tavern, and a community gallery where members publish, collect and remix each other's work. The homepage describes itself as an "AI-enabled creative platform" and leads with Tavern — "interactive roleplay, meaningful conversations, and endless adventures with your favorite anime characters" — before the generation tools. Around that core sit video generation, music generation, voice, a Model Hub, image presets and LoRA training.
The generation side is built on open diffusion model lineages rather than a single proprietary model. The site states its Model Hub spans "From Stable Diffusion to Flux and Illustrious-XL" and claims a catalogue of 105,000+ models and Spells, with official documentation adding dedicated guides for PonyXL and Flux. Users can upload their own checkpoints and train LoRAs to keep a character consistent across images. On the chat side, Tavern offers five in-house fine-tuned models alongside third-party LLMs the site names as GLM, DeepSeek and Claude.
Two facts about identity matter before you go further. First, the operator is YDY AI Limited, a Hong Kong company whose privacy policy lists a Kowloon Bay address and which the terms of service footer identifies as a subsidiary of SYR CORP. Second — and this trips up a lot of write-ups — Yodayo and Moescape are two brand shells over the same product. Their homepages are word-for-word identical apart from the brand name and the assistant's name (Yoda-chan versus Moe-chan), the currency is called YoBeans on one and Mochi on the other with identical earning rules, and the Yodayo buy page tells you outright to purchase credits through the Moescape app.
This is the most developed part of the platform. Beyond prompt-to-image, the official documentation walks through prompt weighting syntax using parentheses and brackets, an improved-prompt helper, model and Spell selection, generation settings, Regional Prompter with ADetailer, an upscaler, HiRes.Fix, image-to-image, ControlNet, and a set of additional image tools. That is a control surface closer to a self-hosted Stable Diffusion workflow than to a one-box consumer generator.
Tavern is not a thin chat wrapper. Its documentation covers manual character creation and an assisted path called Moepilot, Lorebooks and a Lorebook Hub for world context, scan depth support, model selection, tunable parameters, a system prompt, a user persona, a Memory Box, group chats, in-chat image generation and character presets. The five in-house models are published with their sizes and context ceilings: Nephra 8B and Nephra 70B at 8192 tokens (Llama 3 trained on roleplay datasets), Lunara 12B at 8192 (a Mistral Nemo model), Apollo 13B at 4096, and Seth 70B in beta at 8192 (fine-tuned from Llama 3.1 on human creative and roleplay data). The company states all of its exclusive models are trained or fine-tuned specifically for roleplay on Tavern.
The Model Hub, marked Beta in the documentation, lets users upload their own models rather than only consuming a curated list. Combined with LoRA training — "train a Lora/Spell for consistent character" — this is the feature that separates the platform from generators where your style options are whatever the vendor shipped.
The platform has expanded beyond stills. Video generation is a first-class homepage entry naming Minimax, Google Veo 3 and Kling among its models; music generation turns lyrics into songs; voice generation produces speech, and Tavern supports voice cloning for characters. These are real features with documentation, but they are configured around the anime-creation use case rather than as standalone studio tools.
Contribution is paid for in platform currency. Tavern chatbot creators earn a 20% commission on user messages once a public bot passes 4,000 interactions in the trailing 90 days, credited daily, with the creator's own messages on their own bots excluded. Separately, milestones pay 150 beans per 1,000 post likes and collects, 200 per 1,000 model uses in image generation, and 100 per 1,000 Spell uses — capped at 1,000 beans over any 30-day window, though earned beans do not expire.
Building a consistent original character. Train a LoRA on your character, generate a visual set with it, then import that character into Tavern with a Lorebook so the roleplay and the artwork share one canon.
Anime fan art and style exploration. Pull from a large community model catalogue to try styles you would otherwise need to install and configure locally.
Long-form roleplay and interactive fiction. Use Lorebooks, a system prompt, a user persona and the Memory Box to keep a long story coherent, choosing a model by its context ceiling.
Publishing to an audience. Post to the gallery, accumulate likes and collects, and convert engagement into beans through the milestone system.
Running a character as a small product. Publish a Tavern bot, grow it past the 4,000-interaction threshold and take a share of the message revenue it generates.
Adding motion or sound to an existing character. Extend a generated character into short video, give it a cloned voice, or generate a track from your own lyrics.
Anime artists and fan creators who want a large model catalogue plus real control tools without setting up a local diffusion stack.
Roleplay and interactive-fiction writers who need structured world memory rather than a generic chatbot, and who care about which model and context window they get.
Character creators looking for an audience and a small revenue share, given the bot commission and milestone systems.
Model and LoRA trainers who want somewhere to publish their work and see it used.
It is a poor fit for commercial studio work where you need clear rights over your outputs — the licence granted to the platform is broad and perpetual — and for anyone under 18, since the terms require users to be at least 18. It is also not the right platform if you are looking for explicit adult content: the content policy prohibits it outright, which is covered in detail below.
Yodayo is primarily a browser product. The web app requires JavaScript and is served behind Cloudflare, which challenges non-browser clients — a detail worth knowing if you plan to script anything against it.
Mobile is where the distribution gets unusual. The site offers a mobile app, but the download button provides an APK file directly rather than linking to an app store. Checks against both major stores found no listing for this product: an Apple App Store search returns other companion and character-chat apps rather than this one, and the expected Google Play package returns a 404. Sideloading an APK means no store review, no store-level refund path, and — relevant to the content question — no app store age rating for this product to point to.
The relationship between yodayo.com and moescape.ai is the other platform fact that matters. They serve the same product under two brands from the same operator, with the currency renamed between them. Purchases specifically route through Moescape. If you are deciding which domain to use, note that they do not publish identical rules — see the limitations section.
The model is prepaid credits, called YoBeans on this domain, rather than a subscription with feature tiers. Packs observed on the buy page, showing base amount, promotional amount and bonus:
Beans are consumed by high-priority image generation, HiRes.Fix quality enhancement, ControlNet, high-quality video generation, and chatting with the higher-quality language models. Failed image generations do not deduct beans.
The free path is genuine but modest: 100 beans on signup plus 50 claimable daily, expiring 24 hours after claim. Discord quests and weekly contests pay more, as do the creator milestones described earlier.
Two things to note. Purchases must go through the Moescape app — the buy page says so twice. And refunds are described as "only applicable under exceptional circumstances," pointing to a separate refund policy, so treat bean purchases as non-refundable in practice. The buy page also carries a "Monthly" tab alongside "One Time," but no separate subscription tier with distinct features was visible; a URL parameter for subscriptions returned the same credit-pack page. (To be verified — confirm current pack pricing and any subscription option directly on the site.)
Civitai is the closest comparison for the model-sharing and LoRA side, with a much larger general catalogue but without an integrated roleplay chat. PixAI targets the same anime-generation niche with a similar credit model. NovelAI offers anime image generation paired with text generation, on a subscription rather than credits.
On the roleplay side, Character.AI is the mass-market reference point with far more traffic but no image generation and tighter content restrictions. Janitor AI and similar character-chat sites compete for the roleplay audience specifically. SillyTavern — whose name the Tavern feature echoes — is the self-hosted option, giving full control and any model you can connect, at the cost of running it yourself.
For pure generation control with no community layer, a local Stable Diffusion or ComfyUI setup delivers the same underlying models without credits, rate limits or content policy, but requires hardware and setup effort.
Content is restricted to non-explicit material, and this is more restrictive than the platform's reputation suggests. The content policy on this domain classifies everything as PG, PG-13 or R and states plainly that "Explicit content is not allowed." The R tier is defined as "Revealing clothing, violence and gore" — not nudity or sex. The policy is explicit that "any sexually explicit content that contain nudity or excessively lewd content are not allowed for both generation and publishing," that any explicit-rated art, characters or models are banned, and that users are blocked from generating such content in the first place. Nudity is also barred from profile and banner images, with a 48-hour window to replace a flagged image before moderators do it and possible suspension for reposting.
The banned-content list is long and enforcement is termination-level. Prohibited outright, whether fictional, anime-styled or realistic: suggestive content portraying minors, bestiality, non-consensual scenarios and incest, filthy or disgusting words and pictures, deepfakes of celebrities whether or not AI was used, animal cruelty, content promoting hate or violence or discrimination or terrorism or harassment, and self-harm or suicide material. The minor-protection rules go further than a single prohibition, specifying that clothing on minors must be age-appropriate and conservative, that photorealistic suggestive images which could be mistaken for real minors are barred, that poses must be natural and age-appropriate, and that interactions involving minors must be non-sexual. Tavern characters carry two additional bans: underage characters in suggestive or explicit sexual context, and characters based on real-life celebrities, with any realistic imagery required to be AI-generated.
The R tier is opt-in and its visibility is limited by default. Users can opt in to R-rated content, but for viewers with the "Show Sensitive Content" toggle off, R-rated art and characters surface only through followed accounts that have the toggle on, and R-rated models are visible at moderator discretion. So the sensitive layer is neither on by default nor unrestricted when enabled.
The two brands publish different content policies. This is the finding most likely to catch you out. The Yodayo policy is dated 21 August 2024 and has three tiers with an R level plus a whole "Sensitive Content" section. The Moescape policy is dated 9 September 2024, lists only PG and PG-13 — no R tier at all — replaces the wording with "Obscene, vulgar and pornographic content are not allowed," and drops the sensitive-content section entirely. Same product, same operator, two rulebooks of different strictness depending on the domain.
The age requirement is stated but not enforced. The terms say twice that users must be 18 or older — "The Company does not permit those under 18 to use the Service" — and additionally require that all characters in creative work be over 18. But the signup form asks only for name, optional username, email and password. There is no date-of-birth field, no age checkbox and no verification step of any kind. The 18+ rule exists as an assertion in a document, with nothing in the registration flow to back it.
The legal documents disagree with each other on age. While the terms set an 18 threshold, the privacy policy's children's-data clause says only that the company does "not knowingly collect data from children under 13" — the standard COPPA figure. Both numbers sit in the same set of documents without reconciliation.
The terms hedge where the content policy is absolute. The content policy says explicit content is not allowed. The terms say the company "does not allow sharing of obscene/sexually explicit content in general," then immediately note that outputs are AI-generated, that the technology "does not always work as expected," and that no guarantees are made about the suitability of outputs. That is a disclaimer for generation-side leakage rather than a promise that it cannot happen.
Third-party descriptions of this platform are outdated and contradictory. A pixiv Encyclopedia entry on YodayoAI — last updated roughly 591 days before this check — states that "Moescape maintains a policy against NSFW content, [while] Yodayo still allows users to create AI-generated images, including NSFW material." Against the current published policy, that is wrong. Multiple affiliate review sites also circulate a claim that explicit content was removed in August 2025 under payment-processor pressure; no corresponding statement appears anywhere in the site's own materials, and those sources have no verifiable provenance, so that account is not repeated here as fact.
The content licence you grant is unusually broad. Using the service grants the company and its successors "a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license" to reproduce, make derivative works of, publicly display, publicly perform, sublicense and distribute your text and image prompts and the assets produced at your direction — and this licence "survives termination of this Agreement by any party, for any reason." Public posting is the default and posted work is remixable by others.
Liability is capped very low. Total company liability and your sole remedy are limited to what you actually paid through the service, or $100 if you have never purchased anything.
Independent quality signals are thin. The Trustpilot profile for yodayo.com shows 2.8 from just 6 reviews, unclaimed by the company, with the platform noting no history of soliciting reviews. Six reviews cannot characterise a platform of this size, so no quality conclusion is drawn from it here — and with no app store listing, there is no store rating to cross-check against either. No coverage from established technology press was found; search results for this product are dominated by affiliate review sites, several of which operate competing products in the same niche.
Some policy documents have not been updated in a long time. The content policy carries an August 2024 date and the documentation FAQ a September 2024 one, both over a year old at the time of checking, on a product whose feature set has clearly moved on since.
There is a real free tier. New accounts receive 100 beans plus 50 claimable daily, and failed generations do not cost anything. The catch is that daily free beans expire 24 hours after being claimed, and the more expensive operations — high-priority generation, HiRes.Fix, ControlNet, video, and the better chat models — consume beans faster than the free allowance replenishes. Paid packs run from $10 to $500.
Not according to its published content policy. Ratings top out at R — defined as revealing clothing, violence and gore — and the policy states "Explicit content is not allowed," barring sexually explicit content containing nudity or excessively lewd material from both generation and publishing, with explicit-rated art, characters and models banned. Note two complications: the terms of service use softer wording ("does not allow... in general") and disclaim that AI output does not always behave as expected; and the sibling Moescape domain publishes a stricter policy with no R tier at all.
The terms require users to be at least 18 and to meet the minimum digital-consent age in their country, and also require that characters in creative work be over 18. Enforcement is another matter: the signup form collects only name, optional username, email and password, with no date of birth, no age confirmation checkbox and no verification. Separately, the privacy policy states only that the company does not knowingly collect data from children under 13, which does not match the 18 figure in the terms.
They are two brand shells over the same product from the same operator, YDY AI Limited. The homepages are identical apart from the brand and assistant names, the documentation site is shared, and the currency is called YoBeans on one domain and Mochi on the other with identical earning rules. Purchases route through the Moescape app regardless of which domain you started on. Importantly, they do not publish identical content policies.
For images, open diffusion lineages — the site names Stable Diffusion, Flux and Illustrious-XL, with documentation guides for PonyXL and Flux, and a hub the company says holds 105,000+ models and Spells. For roleplay chat, five in-house fine-tuned models with published sizes and context limits: Nephra 8B, Nephra 70B, Lunara 12B and Seth 70B (beta) at 8192 tokens each, and Apollo 13B at 4096. The site additionally names GLM, DeepSeek and Claude as third-party LLMs behind Tavern, and Minimax, Google Veo 3 and Kling for video.
Read the licence clause before assuming anything. Using the service grants the company a perpetual, worldwide, non-exclusive, sublicensable, royalty-free and irrevocable licence over your prompts and generated assets, surviving termination for any reason. Your images are publicly viewable and remixable by default, with a private plan able to bypass some public-sharing defaults. For any commercial use with rights implications, get the terms reviewed rather than relying on a summary.
There is a mobile app, but distribution is by direct APK download from the site rather than through an app store. No listing was found on either the Apple App Store or Google Play. That means no store review, no store-mediated refunds, and no app store age rating for the product.
An automated AI moderator screens uploads and can flag content as BANNED outright, after which human staff — the policy calls them "content police" — review and either approve with a rating or remove it. Sensitive content may be routed to human review, and approved content can still receive limited visibility depending on its rating. Removed artwork, Tavern characters and models can be appealed by email to the moderator address, and every post, model and character carries a report button.
Two mechanisms. Tavern bot creators receive a 20% commission on user messages once a public bot exceeds 4,000 interactions in the trailing 90 days, credited daily, with the creator's own messages excluded. Milestones pay 150 beans per 1,000 post likes and collects, 200 per 1,000 model uses, and 100 per 1,000 Spell uses. Milestone earnings are capped at 1,000 beans in any 30-day window but do not expire. There are also monthly Top Creator bonuses and an Ambassador programme.
The operator is YDY AI Limited, with an address in Kowloon Bay, Hong Kong, identified in the terms footer as a subsidiary of SYR CORP. The privacy policy names the company as data controller for GDPR purposes and references rights under Hong Kong's PDPO alongside GDPR and CCPA. The terms also reserve the right to report suspected unlawful activity to law enforcement and regulators and to disclose profile details, email addresses, usage history, posted materials, IP addresses and traffic information as it deems appropriate.
