Everyday Helper Chat

Status: Live.

1. Your muse

   I want to build a chatbot that can answer anything and utilize any meta tool

2. Show thinking

3. Checks passed

4. 

   The plan: Everyday Helper Chat

   Ask anything by text, voice, or photo and get answers plus images.

   Can build part of it

   1. Open the chat on web or phone
   2. Type a question, record a short voice note, or add a photo
   3. Pick make image or find objects when you want a picture or outline

   Won't do: Answers can be incomplete or out of date

   See the full plan

   Who it's for

   Anyone who wants one simple chat for questions, photos, voice, and image making.

   How it works

   1. Open the chat on web or phone.
   2. Type a question, record a short voice note, or add a photo.
   3. Pick make image or find objects when you want a picture or outline.
   4. Read the answer, view images inline, and follow up.

   Screens

   • Chat. Ask by text, voice, or photo and read answers with inline images.
   • Studio. Generate images, edit an uploaded photo, and view object outlines.
   • History. Reopen past chats saved in the browser.

   What goes in

   • Text: A question or follow up, a few sentences to a few paragraphs.
   • Image (optional): Optional single photo up to 4.5 MB for vision answers, outlining, or edit base.
   • Audio (optional): Optional short voice recording, converted in browser to mono 16-bit WAV before transcription.

   What comes out

   • Text: Plain text answer, with note about sources when web search was used.
   • Image: Generated or edited JPEG shown inline with download.

   Models

   Model	Used for	Why
   Muse Wonder	Answer questions, read photos and PDFs, and run direct web search.	It is the core reasoning and vision model for open ended chat and lookup.
   Muse Image	Generate and edit images from text prompts.	It makes new images and edits a user photo inside the chat.
   Voice Transcribe	Turn recorded voice memos into text.	It lets people speak instead of typing on phone.
   SAM 3.1	Find and outline objects in a photo from a short noun phrase.	It adds boxes and masks for what is where in an uploaded photo.

   Limits

   • Answers can be incomplete or out of date. An answer with no sources says so.
   • Voice replies use the browser built in speech synthesis.
   • Speech input works from recorded files only, not live streams, and returns one plain transcript.
   • Image edits change the whole frame or a cropped region pasted back, not exact pixels.
   • Object outlines are estimates and can miss hidden or small objects.
   • Structured nearby places, events, and paper records are not included in this web and phone version.
   • History stays in the browser only. No accounts and no other services or keys.
   • One run should cost cents. Heavy image or long chat use can hit the daily budget.

   From your inspo

   The person gave a text only note. They want a general chatbot that can answer anything and use Meta models and tools.

   Takes

   • General question answering job
   • Simple chat flow with text, photo, and voice input
   • Inline generated images and plain answers

   Cost per run: about $0.03 (est.). One text turn uses about 1,500 Wonder input tokens ($0.006) plus 400 output tokens ($0.008) plus 1 web search ($0.0025). Image turns add one muse-image-1.0 image ($0.01). Typical mixed run totals about $0.025.

6. Done. Everyday Helper Chat is built and npm run verify passes all six steps.

   It's live.

   musemuse-everyday-helper-chat.vercel.app (opens in a new tab)

   Product pageNew muse

   13 m 50 s · Input 4.8M · Output 43.4k · $0.19 est.

7. Tested what this product comes with

   $0.14 est.

   Show steps