An AI meeting assistant for non-native English speakers

The real problem: time, not vocabulary

Many people who work in a second language know the words. What's hard is the speed. Native speakers talk quickly, cut each other off, use idioms and abbreviations, and expect an answer the moment they stop. By the time you've understood a question and built a sentence, the conversation has moved on, or you answer something slightly different from what was asked.

Co-Help is designed for those few seconds. It doesn't speak for you. It shows you what was said and offers a starting point for your reply, so more of your attention goes to the content.

What you see during the call

• A live transcript of both sides. Co-Help captures your computer's system audio (no bot joins the meeting) and writes down what other people say and what you say in real time. When a question is long, fast or accented, you can read it again instead of asking someone to repeat it.
• Suggested phrasing for your reply. When Co-Help detects a question, it drafts a response on the overlay. Use it as a structure: pick the points that are true for you and say them in your own words.
• A plain-language setting. An audience toggle switches suggestions from technical detail to plain, jargon-free wording. It was built for explaining to non-technical listeners, and it also gives you simpler sentences that are easier to say naturally.
• Memory of the conversation. Co-Help tracks what's already been said, so a follow-up like “and what about the timeline?” gets a reply that fits the earlier discussion.

Consent. Some places require every participant's consent before a call is transcribed. You're responsible for getting it.

Prepare with your own words

Suggestions are better when Co-Help knows your work. Before the call, the setup wizard takes your role and a background document (PDF, Word .docx or .txt), extra documents such as a project brief or product sheet, the other company's website, and free-form notes. Notes are a good place for the names of people on the call, the agenda, and the English terms your team uses for your product, so suggested replies use the same vocabulary your colleagues do.

Rehearse out loud beforehand

Co-Help includes a practice mode. It is designed as a mock interview, so it fits job interviews in English most directly, but it's also useful rehearsal for explaining your work. In the live voice option, an AI interviewer reads each question aloud, waits for your spoken answer, and may ask one short follow-up. Each answer is scored on five criteria, including communication clarity for behavioral questions, with written feedback and a complete model answer that shows one natural way to phrase it. Practice with mock interviews and feedback, prep with your resume and job description, then use the same assistant on the real call. The AI mock interview page explains the details.

After the call

When you end a session, Co-Help shows a session report with the full conversation log, which you can download as an HTML file. Reading back the questions you found hard, and how they were phrased, is a practical way to prepare for the next meeting. After the session the conversation is deleted from the app, so download it if you want to keep it.

Honest limits

• No translation. Co-Help doesn't translate speech or text between languages. It's built for calls held in English.
• Transcripts aren't perfect. Transcription runs on Groq Whisper with OpenAI as fallback. Poor audio, people talking over each other, strong accents and specialized terms can all cause mistakes.
• Suggestions take a moment. A reply is drafted after the question is heard, so there's a short delay. It helps most with questions you have a second to think about, not rapid small talk.
• It can be wrong. A suggestion may misread the question or state something that isn't true for you. You stay the one who decides what to say.
• Desktop only. Co-Help runs on Windows 10/11 and macOS 12+, not on phones.

Pricing and related guides

As of November 12, 2026: start with a free 15-minute trial (no card). Pro is $10 per session, 5 for $39, $79 for 30 days or $349 a year, with sessions up to 90 minutes. Pro Max adds the most advanced models, screenshots and diagrams, and 2-hour sessions: $20 per session, 5 for $75, $150 for 30 days or $600 a year. Every plan includes 20+ languages. See the pricing page.

Frequently asked questions