generator: https://perchance.org/clean-text2image

disclaimer: I'm the generator maintainer.

---

Yeap. You can now generate the 8-bit like pixel arts at my generator now. It is available under the Art Direction. This should accelerate the artwork post-processing work by working on top of an image rather than drawing everything from scratch.

Please, enjoy!

---

This came from a user query about the FLUX.1 Schnell #AI #T2I #aiimage realism quality from Perchance.org. Having worked with SDXL and this, I would say Flux way easier and are very close for deep-faking.

Make yours here: https://perchance.org/clean-text2image

Disclaimer: I'm the #genAI maintainer.

• Art direction: realistic
• Camera Position: High Angle
• Lighting: Cinematic
• Contain Humanoid body: YES
• Guidance Scale: 6
• 18+: NO
• Positive prompt (from a user): A woman with long light brown hair standing on top of a mountain. She is boho chic, carefree.
• Negative prompt: (leave it empty)

Generator: https://perchance.org/clean-text2image

Disclaimer: I'm the maintainer

---

A 3D hand-made cartoonish art direction is now available. You can generate some ideas and craft your own clay models out.

Also, I added a few updates to the generator:

v2.7.0

1. Added Claymation art direction.
2. Enabled programmable variables for Seed Number field.
3. Removed seed vs image quantities restriction for horizontal explorations.
4. Updated programmable tips wordings to reflect the latest changes.
5. Updated SEO description.
6. Corrected 18+ arousal -> arousal typo.
7. Improved High Angle shot prompt assistance sentences for consistency.
8. Improved Low Angle shot prompt assistance sentences for consistency.
9. Improved Portrait shot prompt assistance sentences for consistency.
10. Added Passport camera position.
11. Added Close Up camera position.
12. Added Point of View camera position.
13. Added Dutch Angle camera position.
14. Reset Normal camera position to no string.
15. Added Frontial lighting.

Enjoy!

---

Generator: https://perchance.org/clean-text2text

Disclaimer: I'm the maintainer.

---

Updated the generator to match my latest use cases as a generic LLM. Some updates from previous would be:

1. Download/Clear/Upload transcripts.
2. Added FAQ tab.
3. Added Release Notes (Update) tab.
4. Added Fireside Chats tab.
5. Added Support Maintainers tab.
6. Responsive UI for mobile devices.

Some MISC Observations

I'm still getting back-end AI calls from multiple AI models per message ranging from:

1. Claude 3.5 model
2. ChatGPT 4 model (new discovery today)
3. Deepseek R1 model
4. Custom Anthropic experimental AI model

So far there is no consistency or selection available.

Guard Rails

DEFINITELY GUARDED by back-end model. How to test: try asking it to yield a drug/bomb recipe and all AI models will reject. If it's unfiltered or unguarded, you shouldn't need to jailbreak your prompt.

Some model (not sure which one) will straight up reject NSFW.

---

Generator: https://perchance.org/clean-text2text

Disclaimer: I'm the generator creator and now transitioning to maintainer.

---

All right. I know a lot of you needed help on the T2T side of things. Here, I created a clean T2T generator which is basically a LLM chat to me (zzz).

The goal is for all other T2T generators' creators to debug your generators' problems. Note that I haven't encounter any erratic behavior described in both Reddit and Lemmy forums so far. Just drop me a message at my clean-T2I and I wlll attend during free time. Below are some problems I witnessed but does not go into unusable states.

1. Slow - Yeah. The AI replies are slow compared to the original providers.
2. Multiple AI models - My queries happened to be answered by multiple AI models: Claude Sonnet 3.5, Deepseek R1, and 1 weird custom AI model from Anthropic during my testing. No AI model selection available. Truly perchance.
3. DO NOT feed empty instruction prompt - The AI will spin all kinds of weird stuffs randomly and chaotically. Make sure you do an empty ('') or (undefined) checks before calling the back-end AI.

I'm poor at fictional novel writing etc. I'm a very technical geek who reads scientific white papers daily. It's better you shed some light and guide me to your problem.

Features

For this generator:

1. It is a clean 1:1 chat with no prompt modifications, filters and etc. Useful for debugging.
2. Context feedback and aware (old messages are being feedback for better context so be careful with total conversation length).
3. No markdown rendering yet (need to hunt an interpreter and avoid XSS security attack).
4. Need a new seesion? - refresh the page.

Cheers and enjoy!

Side-notes

I'll be cooling down my involvement with Perchance. Last 2 weeks' new features announcements were receiving weaker and weaker responses and today someone attacked me on my T2I generator. Need some time to cool off.

---

Generator: https://perchance.org/clean-text2image

Disclaimer: I'm the generator maintainer.

---

At long last, the (paper cut) collage art direction is now available for folks to use at my generator. You can use it to generate some visual ideas to produce your actual paper cut collage art piece.

Cheers and enjoy!

---