Published October 5, 2026
Uncensored AI is back on Secret Chat AI, and video generation is now live. The first days of October brought a new Image Lax entry, two video generators, an uncensored chat model, a rebuilt Upscaler and a simpler way to pay. These changes came from a plain problem: creators kept running into refusals while doing lawful work.
We wanted the app to make room for dark fiction, satire, anatomy, political work and difficult visual ideas without pretending that every request should pass. That meant restoring the Uncensored level, marking it clearly and keeping a safety floor across the whole service.
Image Lax Brings the Uncensored Image Generator Back
On October 2, we split image generation into two entries. Image keeps the Standard level and offers GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, Nano Banana 2, Nano Banana Pro and Grok Imagine 2. The uncensored Image generator is listed as Image Lax, with a flame. It offers Seedream 4.5, Seedream 5 Pro, Qwen Image 3 Pro, Lustify and WAI Illustrious.
The split makes the choice visible before you send a prompt. Use the Standard entry for everyday pictures. Choose Image Lax when a lawful prompt or edit is likely to meet a stricter screen. Each provider still applies its own rules, so a picture can still be refused. A failed or refused generation costs nothing.
The controls changed too. Model, number of pictures, shape and resolution now sit in chips under the message box, with the price shown before you start. You can create several variants, attach pictures for an edit and download the result from the chat. WAI Illustrious joined Image Lax on October 3 for anime-style work. Seedream moved to US-hosted Atlas Cloud as its first provider on October 5, with Venice as a backup.
Read the full guide to the uncensored AI image generator, or compare both entries on the main AI image generator page.
Video and Video Lax Are Now Live
Video generation went live on October 4. The Video entry uses the Standard level and includes Kling 2.6 Pro, Kling 3.0 Standard, Kling 3.0 Pro, Veo 3.1 Fast and Grok Imagine Video. Video Lax carries the flame and uses the Uncensored level. Its models are Wan 3.0, Wan 3.0 Prime, Seedance 2.5, Seedance 2.0 Fast and MiniMax H3 Max.
Both entries handle text-to-video and image-to-video. Write a prompt with no picture to start from text, or add a picture to the First frame slot to begin the clip there. Models that support a Last frame can also finish on a second picture. Kling 3.0, Veo, Wan, Seedance and MiniMax accept a last frame; Kling 2.6 Pro and Grok do not.
The settings depend on the model. Across the lineup, clips can run for up to 30 seconds, Veo can reach 4K, and most models offer a sound switch. The price of the complete clip appears before rendering begins. If the provider refuses the request, the download fails or the two-hour wait runs out, the held amount returns to your balance.
On October 5, we added the I can wait control for Kling, Wan and Seedance. It is on by default, so the app keeps waiting for the main provider. Switch it off and, if the clip is still not ready after five minutes, a backup provider starts another render. Both clips may arrive and you pay for each one; the backup clip is charged at the backup provider's own price. You can change the switch while a clip is rendering.
The finished video returns to the chat with a Download button. It also appears in the Videos section of the browser-based Library. See the complete AI video generator guide or the separate uncensored AI video generator page.
The Upscaler Now Handles Small and Large Pictures Differently
The Upscaler was rebuilt on October 3. It is a model choice inside both image entries. Pick it and attach up to 10 pictures; it starts without a prompt.
A picture under 2,048 pixels on its long side is restored and redrawn at about 2K. This route is meant for small, blurry or damaged images, though an AI redraw can alter fine text or a face. A picture that is already 2,048 pixels or larger is enlarged by two without being redrawn, up to the output cap. The price is shown before you start, and a failed picture costs nothing.
GLM Abliterated Replaced Venice in Chat
The text model briefly called Venice became GLM Abliterated on October 3. It is Abliteration AI's uncensored, fine-tuned GLM 5.3, served through Abliteration AI's own API. It offers Smart Agentic, Fast Chat, Web Search, Research and Create Image modes. It can read pictures and documents; PDFs reach it as extracted text.
GLM Abliterated joins Grok 4.7 and DeepSeek V4 on the Uncensored level for chat. Each has a flame in the model list. ChatGPT, Claude, Gemini and Perplexity remain on the Standard level. You can switch families in the same conversation when one model's answer or limits do not fit the job.
We also added Duet Lax to the AI Council. It sends the same question to GLM Abliterated and Grok, then uses an uncensored referee to compare their answers. It is useful when you want two independent readings of a difficult prompt, but agreement between two models is still evidence rather than proof.
What Uncensored Means Here
The flame identifies the lighter of our two screening levels. In chat, that covers Grok, DeepSeek and GLM Abliterated, plus Duet Lax. In media, it covers Image Lax and Video Lax. It means fewer needless refusals for lawful creative, political, historical and clinical work. It does not mean that providers accept every request.
Every prompt, attached document and frame picture is screened before it reaches a model. Anything sexual involving minors, terrorism and real threats of violence stays blocked on every model. Each model provider also keeps its own rules, so changing the level cannot guarantee an answer or a finished image or clip.
The wider uncensored AI guide explains the two levels and the models behind each flame mark.
Payments Moved to Volet and Prices Now Appear in Dollars
On October 2, we moved payments to Volet and stopped selling subscriptions. Secret Chat AI now uses one prepaid balance in US dollars. Add funds with crypto, or pay by card through a Volet wallet. Packs start at $12, and the balance never expires.
Existing annual and gifted plans keep delivering their monthly batches until they end. They do not renew, and nothing more is charged. New use is pay as you go. The app shows the price before an image or video starts and the exact charge beneath each completed result. Read the current options on the payment page.
Privacy Still Has a Clear Limit
Secret Chat AI stores no chat archive on its servers. History stays in your browser, no profile is built from your chats, and prompts are never used for training. Providers see our service rather than a personal model account.
Every prompt, picture, document and video frame still goes verbatim to a third-party model provider. The OpenAI moderation classifier sees the material too. Remove names and other identifying details before sending them. That part remains your responsibility.
These releases are already in the app. Open the model list to find Image Lax, Video, Video Lax, GLM Abliterated and Duet Lax; every Uncensored entry carries the flame.