# First-run setup What the TalkBar welcome window asks, which models it downloads, how to choose where AI runs, and how to run setup again later. Canonical URL: https://docs.talkbar.app/getting-started/first-run-setup The first time you open TalkBar, a setup window gets it ready for your first session. It has five steps: **Welcome**, **AI**, **Voice**, **Workspace** and **Review**. Most people finish it in a few minutes, plus the time the model downloads take. ## How setup works - Click **Continue** to move to the next step and **Back** to return. You can also click any step you have already visited in the step bar. - If something is missing, **Continue** tells you what to do and highlights the field or download that needs attention. - Setup must be finished before you use TalkBar. If you close the window, TalkBar asks **Quit setup?** Choose **Quit TalkBar** to stop for now, or **Continue setup**. The next time you open TalkBar, setup resumes where you left off with the notice **Setup didn't finish last time. Continue from here.** - Downloads keep running in the background while you fill in the other steps. ## Step 1: Welcome **Welcome to TalkBar.** Your profile stays on this device. | Field | What it's for | | --- | --- | | **Your name** | Required. It is saved to your profile on this device. | | **Job title** | Optional. | | **Get TalkBar ready to record** | Lists the models TalkBar needs before your first session: **Speech recognition**, **Speaker models** and **Built-in AI**, with their download sizes. Click **Download all** to start them all at once. | | **App language** | The language of the TalkBar interface. It changes right away. | | **Theme** | **System**, **Light** or **Dark**. | The **Built-in AI** row has an **Include** switch. Turn it off if you plan to use your own AI provider instead; the card then says the model is skipped and you connect a provider in the next step. The card also shows the free space on the drive. If there isn't enough room, it tells you how much is needed. If a download fails, click **Retry downloads**. New to TalkBar? Click **Take a quick TalkBar tour** at the bottom of the Welcome step. The setup window steps aside while the tour runs and comes back afterwards with what you already typed. ## Step 2: AI **Choose how TalkBar runs AI.** Pick where AI runs and make one model ready. | Option | What it means | | --- | --- | | **Local built-in AI (default)** | Runs on this computer after a one-time model download. No external AI provider is used. | | **External providers** | Connect Ollama, a CLI tool, or your own API key. You can skip local downloads. | | **Both** | Download local models and connect an external provider. You can switch, or restrict AI to local models with Private AI Mode, at any time. | Below the options you set up what you chose: - **Local language models** lets you choose the built-in model. Not sure which one? Click **Model Advisor**. It measures this computer and marks its pick. - **External providers** walks you through connecting one: **Local AI (Ollama)**, **CLI tools** (Claude Code, Codex, Agy, Grok CLI or OpenCode), or **Bring your own API key** (Anthropic, OpenAI, Gemini, Groq and others). - When your default model comes from an external provider, click **Verify it works** once. It sends a small test request and shows **Working** when the provider answers. If the check fails, TalkBar says why, for example **The provider rejected the API key.**, and offers **Fix provider**. To continue, one AI model must be ready: the built-in model downloaded, or a provider set up. If you picked the local option but your default model is an external provider, click **Use built-in AI instead**. Want the background first? **Learn about AI first** starts the *How TalkBar uses AI* tour, then returns you to setup. See [How TalkBar uses AI](/ai/how-talkbar-uses-ai) and [Providers and models](/ai/providers-and-models). ## Step 3: Voice **Set up voice.** The models every session needs, and your microphone. | Section | What to do | | --- | --- | | **Speech recognition** | Turns what is said into the live transcript. Choose the **Live transcription model**; Parakeet V3 is fast and recommended. Then set the **Transcription language**. Some models detect the spoken language automatically; others transcribe only certain languages, and the step tells you which. | | **Speaker models** | Voice activity detection and speaker separation. Needed for every session. | | **Microphone** | Optional. Choose the **Input device** used by default for sessions and dictation, and make a short test recording to check it. If no microphone is found, you can still record system audio. | | **Optional downloads** | Extra models you can add now or later: **Large sessions (more than 4 speakers)**, **Translation** and **Documents** (semantic document search). | The speech and speaker models must finish downloading before you can continue. Click **Download all** if they haven't started. ## Step 4: Workspace **Set up your workspace.** Starter packs and an optional first project. Under **What will you record?**, pick any starter packs that fit. Each one switches on the matching session templates and skills. **Meetings & 1:1s** and **Voice notes & ideas** are selected by default. | Starter pack | For | | --- | --- | | Meetings & 1:1s | Team meetings, calls, and one-on-ones. | | Voice notes & ideas | Solo thinking, dictation, and brainstorms. | | Sales & client calls | Discovery calls, QBRs, objections, follow-ups. | | Interviews & research | Candidate interviews and user research sessions. | | Coaching & 1:1 development | Coaching conversations and development check-ins. | | Podcasts & lectures | Recorded shows, talks, and lectures. | | Legal & diligence | Consultations, risk, and compliance-relevant calls. | | Leadership & board | Investor calls, board syncs, and exec reviews. | **First project (optional)** creates a project to file sessions and notes under. Leave it empty to skip; you can add projects any time. ## Step 5: Review **Review and finish.** This step lists **Your setup** (profile, app language, AI, speech recognition, microphone, starter packs and first project) and lets you pick a **Default template** for new sessions. Click **Change** next to any row to go back and fix it. Below that, each **Required** and **Optional** item shows its status: **Ready**, **Missing**, **Downloading**, **Failed** or **Not verified**. When everything required is ready, you see **Everything required is ready**. Click **Finish setup**. ## After setup TalkBar shows **Welcome to TalkBar** and offers the *Find your way around* tour once. Click **Start tour** to take it now, or **Not now** to skip it. You can start any tour later from **Settings → Help & Tours**. See [Getting around](/getting-started/getting-around#guided-tours). ## Run setup again You can revisit your profile, starter packs and AI setup at any time. Existing settings stay as they are. - **Settings → Help & Tours → Start onboarding again**, or - **Settings → Doctor → Start onboarding again**. When you run it again, setup opens on **Welcome** and can be closed at any time. Starter packs you choose are added; existing templates and skills remain unchanged. You can't start it while a recording is running. To change just the AI setup, click the TalkBar logo and choose **Set up AI…**, or open **Settings → AI Providers** and click **Setup AI**. ## Related - [Install and update](/getting-started/install-and-update) - [Your first session](/getting-started/first-session) - [Local models](/ai/local-models) - [Private Local AI settings](/settings/private-local-ai) - [Voice settings](/settings/voice)