LL AI v64.2.1

Notes

Saved automatically to this chat

No files or notes

0 characters

•••Settings & tools

Ready to test

LL AI · v64.2.1

Ready

What can I help with?

Ask anything, or add files and notes for extra context.

PluginsChoose tools

Tools stay on until you turn them off.

LL can make mistakes. Check important information.

＋

Drop files herePDF, DOCX, text, images, audio, or video

Images

✦ Free image studio

Turn an idea into an image.

Describe what you want, choose a visual direction, and LL will generate it through your connected free provider.

Setup neededPrivate gallery · saved on this device

◇Start with a clear subjectAdd the setting, lighting, viewpoint, and mood for stronger results.

LOCAL GALLERY

Your creations

0 images

No generated images yet
Describe one below to get started.

✦

Free image generation

Choose the free route LL should use. Cloudflare is the best option for a shared public app; Hugging Face is a convenient personal fallback.

Cloudflare image-worker URL

Paste the URL of your deployed Cloudflare image Worker here. To make the URL work automatically for everyone, place it in image-service-config.js before uploading LL AI.

Hugging Face token

Model

Create a fine-grained token with permission to call Inference Providers. The token stays in this browser. Hugging Face’s free credits are small and reset monthly.

✓

Your provider credentials stay in this browser.The Cloudflare Worker URL is public and safe to share. Hugging Face tokens are stored only on this device.

LL Live

Voice mode

Start speaking. LL listens automatically.

Premium voice

ElevenLabs is optional. Choose and verify a premium voice here, or leave it unset and LL will automatically use the best voice available on this device.

ElevenLabs API key

Available voice

Voice ID Speech model

Language intelligence

LL uses Whisper Large v3 with automatic language detection, acoustic confidence checks, and Arabic/English comparison passes only when a short utterance is genuinely ambiguous. It never rewrites your transcript with the chat model.

Automatic multilingual understanding Whisper detects the spoken language automatically. Arabic and Arabizi replies use natural Egyptian Arabic without a manual language switch.

Live transcription engine

Arabic voice (optional)

Use conversation language as a fallbackRecent turns are consulted only when the automatic transcript is low-confidence. Clear speech is never overridden, and code-switching stays intact. Native pronunciation preparationConverts Arabizi into spoken Arabic script, prepares Arabic replies as natural Egyptian spoken Arabic automatically, preserves genuine English words, and adds pronunciation-friendly pauses without changing the visible answer.

99+ languagesEgyptian Arabic repliesCode-switchingConfidence gateNo AI rewriting

Your key stays in this browser. Tap Refresh to verify it and load voices.