I built VoiceForge, a desktop dictation app for macOS, Windows, and Linux. The idea is simple: pay $79 once for the app, then choose how your transcription runs. There is no VoiceForge app subscription.
You can use local Whisper transcription without an API key, or bring your own key for a supported cloud provider, including Groq. Local transcription with cloud AI cleanup switched off has no provider usage bill.
What that means for the cost
Wispr Flow offers a free plan, and its Pro plan currently costs $15 per month or $144 per year. VoiceForge's app license is $79 once, including future updates. Optional cloud transcription and AI cleanup are separate provider costs, so the total depends on the setup you choose. These products have different features; this is a comparison of the license costs, not a claim that they are identical.
Groq has a free tier you can use with your own account and API key, subject to its usage limits. You do not have to upgrade to a paid API plan to start. Free-tier availability and limits can change, and paid cloud usage is billed by the provider.
For developers who are comfortable choosing a provider or downloading a local model, that gives you more control over where transcription runs and what you spend.
A workflow I use: speaking to a coding agent
One workflow I use VoiceForge for is speaking a request directly into my coding tool. The useful part is having room to explain the task. A prompt can cover three things:
- What happened.
- What should happen instead.
- How to check the change, including what should stay as it is.
For example:
The settings panel loses the selected microphone after I restart the app. Find where that selection is saved, preserve it across restarts, and test with two different input devices. Keep the layout as it is.
That is an example prompt, not a claim that the agent already fixed the bug. Before sending a spoken request, I would check the transcript for file names, numbers, and words like “not.” Those small errors can change the task. I still use the keyboard for precise syntax and edits.
Here is a 13-second recording of VoiceForge putting a spoken prompt into Codex.
A practical test: dictate your next five short agent prompts, review each one, and note whether the extra context helped or the corrections slowed you down. You can try this with a dictation tool you already have.
VoiceForge has a 30-day money-back guarantee. I am its creator, and this post was drafted with AI assistance. The linked demo is a recording of my own session.
If you have tried dictating to an AI coding tool, what creates the most friction: technical vocabulary, corrections, the microphone shortcut, or speaking in a shared workspace?
Top comments (0)