Qualifizer — AI Content Studio in Telegram
Post → matching image → edit → done. All in one Telegram chat.
0 apps
image generation and editing without leaving Telegram
context-aware
bot knows your post, photo, and previous tasks
voice → image
describe what you need out loud — get the result
The problem
You write a post, then open Midjourney to generate a matching image — but Midjourney doesn't know the post, so you copy-paste text manually. Image comes out wrong. You edit the prompt, regenerate. Then you need to edit an existing photo — that's a third app. Every tool is an island. The work takes 5 minutes. The switching takes 30.
How it works
Describe what you need
Text, voice message, or reply to an existing post — any input works
Context loaded
Replied to a post? Bot loads its content. Referenced 'that image'? It finds the last relevant task from Supabase
Intent classified
Gemini Flash: is this image_gen or image_edit? Which source context applies?
Prompt enhanced by AI
Your instruction + source context → AI writes a detailed professional generation prompt
Result sent to Telegram
Flux Pro ultra renders and sends directly to your chat. Reply to refine
Image editing — AI sees the photo
Send any photo with an instruction: 'remove the background', 'make it look cinematic', 'add dramatic lighting'. Gemini 2.5 Flash with vision reads the photo and your instruction together — writes a professional editing prompt. NanoBanana API applies the edit. Result lands in Telegram. Reply to keep refining.
The payoff
Knows your post
Reply to any text you wrote — the bot reads it before generating. Image matches the topic, tone, and key message automatically.
Remembers what you refined
Each task stored in Supabase. Reply to any previous output and the bot picks up exactly where you left off.
Voice works too
Record: 'dark moody city at night for my AI finance post.' Whisper transcribes, bot generates.
30 min → 5 min per piece
At 10 content pieces per week — that's 4 hours returned every week. In one Telegram chat.
What's under the hood
n8n — orchestration
Main router workflow (56 nodes) + Image Gen Agent + Image Edit Agent as separate sub-workflows. Each callable directly.
Gemini 2.5 Flash — router + vision
Routes intent in the main workflow. Used again in Image Edit Agent to analyze photos and write enhanced editing prompts.
Flux Pro ultra — generation
State-of-the-art image model. Prompt enhanced by AI before calling — results are coherent, not random.
Supabase — task memory
Every task stored with full context: source content, prompt used, output. Reply context loaded via Postgres RPC.
Open. Modular. Extendable.
All workflow JSONs are open. The router doesn't care what agents are downstream — add a new specialist by creating a sub-workflow and adding one Switch case.
Context-aware image gen
Reply to your post with 'make an image for this' — the bot reads the post and generates an image that matches the content, not a random prompt.
AI-powered image editing
Send any photo with an edit instruction. AI vision enhances your prompt into a detailed professional directive before calling the editing API.
Voice input
Describe what you want out loud. Voice transcribed via Whisper, processed like text. No typing required.
Iterative refinement
Reply to any generated image to refine it. The bot loads the original context and regenerates with your adjustment.