Wan has a permissive, filter-light reputation in the open-weight AI art community when self-hosted, but its behavior on Picasso IA's hosted API is untested and unconfirmed. Try your prompt and see what happens: if Wan generates it, that is your answer for this platform; if it refuses, the app switches you straight to Picasso Video or LTX 2.3 Pro, confirmed filter-free video models, with your prompt carried over. Either way, you can start generating uncensored video today.
If you have spent time in the open-weight AI art community, you have probably heard that Wan is one of the more permissive video models out there when self-hosted, generating adult content with far less resistance than the big closed platforms. That reputation is real, and it is not the thing this page is about. What this page is about is what Wan actually does on Picasso IA, running through our hosted API, and the honest answer is that we have not tested it against adult prompts here and cannot confirm either outcome.
Those are two different questions, and conflating them is where most confusion starts. A model's behavior when someone runs it locally, on their own hardware, tells you about that specific setup. It does not automatically carry over to a hosted version behind an API, because a hosted deployment can layer its own safety system on top of the base model regardless of how permissive the underlying weights are known to be elsewhere. We are not saying Wan is filtered on Picasso IA. We are not saying it is open. We are saying we do not know yet, and we are not going to borrow a reputation earned somewhere else and print it here as fact.
So if you came looking for a guaranteed uncensored video model, Wan is not the page that gives you that guarantee, at least not yet. What it gives you is a model worth testing, with a fast, no-cost way to find out where it lands for your specific prompt, and a clear fallback if it refuses. If your priority is a video model already confirmed to handle adult prompts without a filter on this platform, jump straight to Picasso Video or LTX 2.3 Pro further down this page. If you are curious whether Wan's open-weight reputation holds up here, keep reading, because the rest of this page is about testing it properly rather than guessing.
Open weights mean the model's parameters are published and anyone with the hardware can run them directly, which is exactly why the self-hosted community has pushed Wan in directions its original release notes never promised. Running your own copy on your own machine, you control the whole pipeline, there is no intermediary layer inspecting your prompt before it reaches the model. That is the setup the "filter-light" reputation comes from, and it is a fair reputation for that specific setup.
A hosted API is a different pipeline. When a platform offers a model through an API rather than letting you run the weights yourself, that platform can add its own moderation step in front of or around the model, independent of whatever the base weights would do on their own. Some hosts leave a permissive model essentially as-is. Others wrap even an open, filter-light model in a stricter layer for liability or policy reasons that have nothing to do with what the model is technically capable of generating. Neither choice is visible from the outside until someone sends a prompt through and sees what comes back.
That is the exact gap between reputation and confirmation. Wan's open-weight reputation is a fact about Wan running outside Picasso IA. What Wan does on our hosted API, specifically, is a separate fact that has to be established by testing, not inferred from a community reputation, however well earned that reputation is elsewhere. Until we have run enough adult prompts through Wan on this platform to say something concrete, we are not going to publish a claim in either direction, and neither should you assume one based on what you have read about self-hosted setups.
This is not unique to Wan. Every model in the toolkit sits behind its own filter logic, set by whoever built or hosts it, and that logic can be strict, permissive, or somewhere in between regardless of the model's public reputation. Six models on Picasso IA are confirmed, by our own testing, to generate adult content directly with no filter standing in the way: Seedream 4.5, Qwen Image 2, Recraft V4.1 Pro, and Picasso Image Editor Pro for images, plus Picasso Video and LTX 2.3 Pro for video. Wan is not on that list yet. It may earn a place on it. Today, it has not.
Testing Wan costs you a couple of minutes and, if the app blocks the generation, no credits. The fallback path sits on the same screen as the refusal, so there is no dead end.
If Wan accepts your prompt, you now have a confirmed data point for that specific case, and you can keep working with it. If it refuses, step five gets you moving again immediately with a model we already know handles adult content on this platform.
While Wan's status on Picasso IA remains untested, two video models on the platform are confirmed to generate adult content directly with no safety filter standing between your prompt and the render: Picasso Video and LTX 2.3 Pro. These are the exact models the refusal screen routes you to if Wan blocks your prompt, and a reasonable starting point if you would rather skip testing entirely.
Picasso Video handles straightforward text to video and image to video generation and does not screen out explicit description or adult framing. LTX 2.3 Pro covers similar ground with its own rendering character, so if the first model's output does not match what you had in mind, running the same prompt through the second one is worth the couple of clicks it takes. Both sit in the toolkit next to Wan, so moving between all three costs you nothing but a few taps.
If your project is a still image rather than a moving one, the confirmed list widens to four more models: Seedream 4.5, Qwen Image 2, Recraft V4.1 Pro, and Picasso Image Editor Pro. Between those four and the two video models above, six models across the catalog are confirmed to handle adult prompts without a filter, out of the far larger set you will find in the full catalog. Wan sits in that larger, unconfirmed set for now, alongside models like GPT Image 1, DALL-E 3, Imagen 4, and most of the rest of the video lineup.
This applies to Wan, to the six confirmed models, and to every other model on Picasso IA without exception: adult content generation is limited to fictional characters only. Do not attempt to generate a real person's likeness, a public figure, a celebrity, or anyone recognizable as a specific living individual, in any adult or suggestive context. Do not attempt to generate anything resembling a minor, under any framing or context. These are not per-model settings you can toggle, they are the baseline the platform runs on, regardless of whether a given model's filter happens to be strict or permissive.
A few things that follow from that baseline in practice. Uploading a real photo of a person and asking any model, Wan included, to place them into adult content is against the rules even on a model whose filter technically permits the generation. Age-ambiguous framing, school settings, or childlike proportions are treated as attempts to generate content involving a minor and are never permitted. If you are unsure whether a prompt crosses a line, rewrite it as a clearly adult, clearly fictional scenario before you submit it rather than testing where the boundary sits.
New accounts on Picasso IA start with free credits, enough to run the five-step test above on Wan and, if needed, switch to Picasso Video or LTX 2.3 Pro, without spending anything up front. Video generation typically costs more credits than a still image because it renders many more frames, so keep that in mind if testing several prompts back to back.
Full plan tiers and current credit allowances live on pricing. Across every plan, a generation that a content filter blocks and returns as "Content Not Supported by This Model" does not consume your credits for that attempt. That means testing Wan first and falling back to a confirmed model if it refuses does not cost you twice, whichever plan you are on.
We cannot confirm that either way yet. Wan has a filter-light reputation in the self-hosted open-weight community, but that reputation describes running the model outside Picasso IA. Its behavior on our hosted API has not been tested against adult prompts, so the honest answer is to try your specific prompt and see what happens rather than assume the self-hosted reputation carries over.
Because a hosted API can add its own moderation layer independent of what the underlying open weights would do if you ran them yourself. Running your own copy of Wan puts you in full control of the pipeline with no intermediary layer. Running it through a hosted API means the host can wrap it in stricter or looser moderation for reasons that have nothing to do with the model's technical capability.
You will see an on-screen message that reads "Content Not Supported by This Model." Underneath it, the app offers to switch you to a model confirmed to handle adult prompts, carrying your exact prompt over automatically so you do not have to retype anything.
Two: Picasso Video and LTX 2.3 Pro. Alongside four confirmed image models, Seedream 4.5, Qwen Image 2, Recraft V4.1 Pro, and Picasso Image Editor Pro, that makes six models total confirmed by our own testing. Wan is not currently one of them.
No. A filter evaluates the whole scene rather than isolated words, so a reworded prompt is likely to get blocked again, and it wastes credits on generations that still fail. The approach that actually works is switching to a model already confirmed to accept adult prompts directly, which the app offers you in the same refusal screen.
No. Any generation blocked by a content filter and shown as "Content Not Supported by This Model" does not consume your credits, so testing Wan before switching to a confirmed model costs you nothing extra.
No. Adult content generation on Picasso IA is restricted to fictional characters only, and that rule applies across every model, confirmed or unconfirmed, Wan included. Do not upload or describe a real, identifiable person in an adult context.
No, never, under any framing or styling. This is an absolute rule with no exceptions, applying across the entire platform and not just the models discussed on this page.
No. New accounts receive free credits that cover initial testing, so you can try Wan and, if it refuses, switch to Picasso Video or LTX 2.3 Pro, before deciding whether to subscribe. Full plan and credit details are on pricing.
The full catalog lists every model on the platform, video and image alike, including Wan and the six confirmed adult content models discussed above. The toolkit is where you actually select a model and start generating.
