PoeDictate to Poe Bots, Not a Generic Chat Box

Voice prompts for Poe bots and multibot chats

Speak prompts into Poe's composer for any bot, multibot @-mentions, and custom bot setup. Review text, then send.

Download AICHE
Works on
macOSWindowsLinux

The short answer: open poe.com, pick a bot from the sidebar or start a multibot thread, click the message composer, press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux) to start recording, speak the full prompt for that bot (constraints, context, output format), press the hotkey again to stop, and AICHE inserts formatted text for you to review before you send.

Poe is a bot platform, not a single-model chat window. You choose bots from the model picker, @-mention additional bots in one thread, build prompt-based custom bots, subscribe to creator bots, and share chat links. Each of those flows needs a strong user message. Poe's built-in voice (where available) is for spoken back-and-forth with one bot. It does not reliably land a structured, editable prompt in the composer for multibot workflows or bot setup. AICHE runs cloud-default transcription. Audio streams to Groq and is discarded immediately after processing, within 1 second. Message Ready cleanup drops text into Poe's composer so you edit and send on your terms.

The Problem

Poe rewards long, specific prompts to specialized bots. A coding bot needs language, framework, and failure mode. A writing bot needs audience and format. A multibot thread needs a clear ask per @-mention so each model does useful work. Typing that detail once per bot in a comparison thread is slow, so people shorten prompts and lose what makes Poe worth using.

Bot creation is worse: the system prompt and opening instructions are the product. Creator bots and API bots on the Poe Creator Platform expect paragraphs of behavior, guardrails, and examples. Subscription bots cost points; weak prompts waste quota. Share chats expose your thread to others; vague first messages make shared links useless.

What Changes

Speaking runs about 150 words per minute. Typing runs about 40. A 200-word bot prompt that takes five minutes to type takes about 80 seconds to speak, and you are more likely to keep the sections you would have dropped.

How It Works

  1. Install AICHE on macOS, Windows, or Linux (desktop).
  2. Open Poe in the browser or Poe desktop app and open a chat with any bot, or start a new thread.
  3. Click into the message composer (or a bot-creation prompt field on poe.com/creators).
  4. Press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux) once to start. The hotkey is toggle on/off, not push-to-talk.
  5. Speak your Poe prompt: role, task, format, constraints, and what to do with prior messages in a multibot thread.
  6. Press the hotkey again. AICHE transcribes, applies optional Message Ready cleanup, and inserts at the cursor.
  7. Review the text, edit a line if needed, then send in Poe as you normally would.

Offline: if you lose connectivity, AICHE queues recordings locally and processes them when you are back online. Text inserts when transcription completes.

Mobile: the global hotkey is for desktop. On phone, record in the AICHE app and paste into Poe manually.

Model Picker and One-Bot Chats

The sidebar model picker is how you open GPT, Claude, Gemini, Llama, image bots, and community bots. The workflow is the same for every bot: select it, click the composer, toggle the hotkey, speak a complete user message, stop, review, send.

Name what the bot should optimize for in the first sentence: "You are reviewing Python asyncio code" beats "help with my code." Poe bots often ship with a fixed system prompt; your message still sets the task, files, and output shape. Enable Clean Language in AICHE settings if you want filler removed and lists tightened before insert.

Multibot Chat and @-Mentions

Multibot chat lets you compare and chain bots in one thread. @-mention a bot to pull it into the conversation, run recommended comparisons, or pass context from an earlier reply into the next bot (search bot for facts, then a writing bot for the draft).

Dictation fits multibot because each @-mention deserves its own clear ask. Speak the message for Bot A, send, @-mention Bot B, dictate a follow-up that references what you need from B without retyping the whole thread. Poe also allows sending a new message before the previous bot finishes responding; you can queue the next dictated prompt while a slow reasoning model still streams.

Heads-up: each bot sees the thread context Poe provides. Say explicitly when a message is only for the bot you @-mentioned ("ignore prior image output; only list API risks").

Custom Bots and Bot Creation Prompts

On poe.com/creators, prompt-based bots need a system prompt, greeting, and suggested user messages. API bots need endpoint description, behavior, and user-facing parameter docs. That text is the bot.

Click into the creation form fields, toggle the hotkey, and dictate once:

  • Who the bot is for and what it must refuse
  • Output format (JSON, markdown table, step list)
  • Tools or knowledge boundaries ("no medical diagnosis," "cite sources")
  • Example user messages that teach subscribers how to prompt it

Private bots stay on your profile; public bots are discoverable. Either way, a spoken 90-second spec beats typed bullet fragments that every subscriber inherits.

Subscription Bots, Points, and Creator Bots

Poe subscription unlocks higher message limits and premium bots. Creator and API bots may charge per message or token usage reported from your endpoint. Strong prompts reduce wasted retries.

When testing a paid bot, dictate a full evaluation prompt up front: inputs, expected structure, edge cases, and how to say "insufficient context." You spend points once per thorough test instead of three shallow tries.

Shared subscription (family or team) shares the subscriber's points pool; clear dictated prompts help every member stay within budget.

Share Chats

Poe lets you share a chat link so others see the thread. Shared links expose your prompts and bot choices. Dictate the opening message as if a colleague will read it cold: goal, constraints, and what was already tried.

Before sharing, read the inserted text once. AICHE does not send messages; you control what leaves the composer.

Voice in Poe vs Composer Dictation

Poe voice (in-app) AICHE into composer
Output Spoken or streamed reply Text you edit
Best for Quick Q&A with one bot Long prompts, multibot @-mentions, bot creation
Multibot One bot per voice turn Full text per @-mention, copyable

Common Questions

Q: Does AICHE work in Poe's desktop app and browser?
A: Yes. Click into Poe's message composer or any bot-creation text field, use the hotkey, speak, press again. AICHE inserts into the focused field at the OS level.

Q: Can I dictate the same prompt and paste it to another bot for comparison?
A: Yes. Dictate once, send to the first bot, copy the composer text (or select and copy the inserted block), switch bots or @-mention another, paste, edit if needed, send. Consistent text makes comparisons fair.

Q: Will @-mention bot names transcribe correctly?
A: Add bot handles and model names to Custom Vocabulary in AICHE settings so Claude-style or community-bot mentions spell the way Poe expects.

Q: I'm building an API bot on the Creator Platform. Can I dictate the settings description?
A: Yes. Any text field on poe.com/creators accepts the same hotkey workflow. Dictate pricing notes, parameter help text, and system instructions separately per field.

Q: Does AICHE send the message or spend Poe points?
A: No. You review inserted text and click Send in Poe. Points and subscriptions are unchanged.

Q: Non-English prompts to English-first bots?
A: Turn on Auto-translation in AICHE settings. Speak in your language; English inserts into the composer for bots that expect English user messages.

Result: Poe stops feeling like "type a short question, hope the bot guesses." You run multibot threads, creator setup, and shared chats with prompts long enough to match each bot's job.

Try it now: open Poe, start a multibot thread, @-mention two bots you use for different tasks, dictate a full prompt for the first, send, then dictate a follow-up for the second that says exactly what context to use from the thread.

Tags

ai-codingvoice-commandsproductivity