Record. Compose. Narrate. All local.
A CPU dictation engine, a GPU AI engine, three creative surfaces, 80+ template operations, and hands-free controls to start, stop and copy dictation. Your audio and text stay on your PC, dictation works offline once the model is downloaded, and no graphics card is needed to dictate. Here is how each piece works.
Parakeet v3 + Whisper - the recording brain
Engine one of two. NVIDIA's Parakeet TDT v3 transcribes on your CPU (int8, ~670 MB) - English + 24 European languages, ready in seconds, and on fast natural speech it beat Whisper large-v3 running on a flagship GPU in our published benchmark (2.0 percent word error against 3.1). Whisper (every size up to Large v3 Turbo on an NVIDIA GPU, tiny on the CPU) stays one click away for ~100 languages and all-GPU rigs.
- No GPU needed - NVIDIA Parakeet v3 dictates on any modern 4-core x64 CPU
- Continuous dictation, overlapping chunks with auto-dedup
- Chunk duration tunable (3 / 5 / 15 / 30 s)
- Engine + device switch in two clicks on the Models page
- Auto language detect, or pin one per session
- Custom vocabulary (Pro 25 terms, Studio up to 500) - list the names and jargon it gets wrong and it corrects close mis-hearings in the transcript, on either engine
- Separate engine process over stdio - it goes online only for model weights
Ollama models, 3B to 70B - the cleanup brain
Engine two of two - and the part your NVIDIA GPU is reserved for. Pair the raw transcript with any local LLM Ollama can serve - Gemma 3, Qwen 3, DeepSeek-R1, Llama 4 Scout, Phi-4, Mistral, whatever. 22+ in the catalog, plus custom prompts of your own: 1 on Free, 5 on Pro, 30 on Studio. The orchestrator parses param counts from model tags and estimates VRAM automatically, so 'pull a new model and it just works' is real.
- 17 templates with 80+ tones (dictation, rewrite, chat, email, social, video, meeting notes, coding prompts...)
- Custom system prompts for medical notes, legal briefs, meeting minutes
- Per-app tones in Mini Mode (Studio) - bind an app once and dictating into it uses that template automatically
- Token-level streaming, cancel mid-run, run-count tracking
- Localhost-only (127.0.0.1:11434), no text sent off the PC, verifiable in netstat
Smart VRAM swap - why both engines coexist
Voxmelt reads your card's total VRAM, measures Whisper's footprint and estimates the AI model's from its tag. By default both stay loaded for an instant handoff, and a model too big to fit beside Whisper is refused at load instead of swapped in silently. On a smaller card, turn off Keep both models in VRAM and Voxmelt swaps one model in at a time.
- Live VRAM telemetry, updated every 2 seconds
- Color-coded VRAM fit: green when a model fits easily, red when it is tight, an amber warning when VRAM runs short
- One-click Free GPU - dumps every model so you can game or render
- Idle auto-free after 1 / 5 / 15 / 30 min. Alt-tab to a game, VRAM comes back
Privacy by architecture, not by ToS
Both engines live on your machine. Speech-to-text runs in a separate engine process that talks to the app in JSON over stdio and goes online only for model weights. Ollama is bolted to 127.0.0.1. The app has no usage analytics and no crash reporting, and your audio and text never leave the box. Voxmelt goes online only for the trial and licence check (a hashed device ID), the optional account and purchase, and anonymous update and model downloads, and none of it carries audio or text. The website counts visits with cookieless site statistics and sets no tracking cookies. So there is no Voxmelt-side processor to sign anything with, but your own privacy and legal obligations still apply to what you store, sync and share.
- No API keys. No account required to record
- Works with WiFi off once the models are downloaded
- The speech engine sends no audio or text anywhere; it goes online only for model weights
- Ollama localhost-only - receipts in netstat, not promises
Three creative surfaces built on those engines.
Record is the flagship. Compose turns the same local AI into a text studio, and Voiceover adds a local voice engine for audio - no mic required, no cloud processing.
Compose - the text-first AI workspace.
Not everything starts with your voice. Paste any text - a rough draft, a colleague's email, meeting notes from another tool - pick a template and a tone, and let the local LLM reshape it. Same AI engine as Record, same templates, same tier access. No recording needed, no cloud processing, same local AI. Think of it as a private, local alternative to pasting into ChatGPT - except your text never leaves your machine.
- Paste or type any text - runs through the same 80+ template tones
- Decoupled template + tone selection - mix any action with any style
- Streaming output with copy, re-run, and Read Aloud on the result
- "Use last transcript" shortcut - bridge a recording into a different tone
- Same daily run limits, same tier gating, same model picker as Record
- Full AI Output panel - tone-shift chips, voice edit, and WAV export (Pro)
Voiceover - text to audio, on your CPU.
Turn any script into narration without a recording booth. Four distinct personas (Clara, Marcus, Maya, Sam) crossed with eight delivery tones (Professional, Conversational, Warm, Calm, Bright, Authoritative, Storyteller, Energetic) give you 32 voice combinations - all synthesized locally on your CPU via Kokoro ONNX, no graphics card needed. No per-character billing, no cloud API, no rate limits. Export clean WAV files (Pro) straight to your Downloads folder. Unlike ElevenLabs or Play.ht, your scripts never leave your machine and there is no usage meter ticking.
- 4 personas x 8 tones = 32 voice combinations, all on-device
- Speed control from 0.7x to 1.5x (Pro+) - match any pacing need
- WAV export (Pro) to Downloads - drag into your editor, timeline, or DAW
- Sentence-by-sentence playback tracking with live highlighting
- "Use last AI output" shortcut - polish in Compose, narrate in Voiceover
- No per-character billing - generate as much audio as your PC can handle
Personas
- Clara - clear, articulate female voice (Free)
- Marcus - measured, corporate male voice (Free)
- Maya - warm, expressive female voice (Pro)
- Sam - relaxed, conversational male voice (Pro)
Delivery tones
- Professional, Conversational, Warm (Free)
- Calm, Bright, Authoritative (Pro)
- Storyteller, Energetic (Pro)
- Speed control 0.7x - 1.5x (Pro)
80+ AI operations - one click, every time.
Templates are how Voxmelt turns a raw transcript or pasted text into exactly the format you need. Each template pack contains multiple tones - not just “rewrite” but “rewrite as a casual Slack reply” or “rewrite as a formal decline email” or “rewrite as a LinkedIn post.” Pick the action, pick the style, get the output. Every template works in both Record and Compose.
Quick edit
Vibe Dictation (6 tones), Rewrite (5), Proofread (4)
Tone shift
Diplomatic (5 tones), Persuasive (5)
AI and code prompts
AI Prompt Architect (9 tones), Coding Prompt (8)
Communications
Chat Responder (6), Email (5), Social Post (4), Video Script (4)
Explain and translate
Meeting Recap (4), Notes (4), Translate (6 languages)
Custom
Write your own system prompt. Free: 1 slot, Pro: 5, Studio: 30
Agentic prompt tones - for vibe coders & AI power users
Dictate a rough idea; get back the exact scaffold your coding agent or LLM actually wants. The Coding Prompt and AI Prompt packs now turn a spoken sentence into deterministic state machines, self-verifying execution loops, agent-gated test beds, multi-agent routing, boundary-calibrated few-shot sets, and strict JSON mode - the prompt patterns power users hand-write today, now one tone away and run by the local AI on your own PC.
Coding Prompt
- Loop EngineerPro
Autonomous control loop with an explicit trigger, goal, verification phase, and retry rule.
- State Machine ArchitectStudio
Deterministic state chart - states, transitions, guards, and invariants, no surprise side effects.
- Agentic Test Bed (TDD)Studio
Isolated, assertive tests written as strict pass/fail gates an agent must clear.
- Speculative RefactorStudio
Two or three refactor paths compared on complexity, performance, and overhead.
AI Prompt
- Extended Deliberation LoopPro
Forces step-by-step reasoning and a self-check before the model commits to an answer.
- Zero-Prose JSON SchemaPro
Locks output to strict JSON - schema in, zero prose or markdown out.
- Cognitive Multi-Agent RouterStudio
Splits the job across Planner, Critic, and Executor blocks with explicit routing.
- Few-Shot Label BoundaryStudio
Balanced edge-case examples that calibrate a small local model right on the decision boundary.
Vibe Dictation
Polish raw spoken thoughts, rants, or logic into structured text in any vibe.
Chat Responder
Turn raw spoken reactions into perfectly vibed replies for Slack, Teams, or text.
AI prompt
Convert spoken thought into a clean, structured LLM prompt.
Coding prompt
Developer-ready spec for AI coding assistants.
Meeting Recap
Turn a recorded meeting, call, or standup into action items, decisions, minutes, or a TL;DR.
Distill Notes
Compress a long voice memo, lecture, or brain-dump into bullets, an outline, a TL;DR, or study notes.
Signal Extractor
Strip away all the noise and keep only the dense nucleus - a one-line gist, a pure deliverables list, or a single atomic note.
Translate
DefaultTranslate the transcript to another language.
Custom prompt
Pick a persona, add your own instructions - runs 100% on your GPU, like every built-in.
Turn a spoken intent into a ready-to-send email - replies, follow-ups, declines, outreach, apologies.
Social post
Shape a raw thought into a post that lands - a LinkedIn post, an X thread, a hook, or a caption.
Rewrite
The everyday rewriter - paraphrase, shorten, expand, formalize, or sharpen any text.
Proofread
Fix and tighten - grammar, clarity, conciseness, and direct active voice.
Diplomatic
Recast a blunt or risky message so it lands well - tactful, calm, and relationship-safe, without losing the point.
Persuasive
Recast flat text into something compelling - confident, benefit-led, and hard to say no to, with the meaning intact.
Explain
Make any idea click - explain it simply, step by step, by analogy, or with a worked example.
Video script
Spoken idea to a script you can read on camera or send straight to the Voiceover page.
Two ways to drive it without touching the keyboard.
The engines do the thinking. These are how you run them - eyes off, hands free, never breaking flow.
No-Hands Mode - run dictation by voice.
Once you switch it on in Settings, a second, always-on Whisper tiny model (about 0.3 GB, on your NVIDIA card if you have one, else on the CPU) listens in the background and fuzzy-matches your speech against short, phonetically-distinct phrases. Say "go ahead" to start dictating, "wrap up" to stop, "copy that" to grab the clean text, "warm up" to load the AI - no hotkey, no mouse, no looking at the window. The listener steps aside the instant real dictation starts so it never fights your main model for the mic or the GPU, then resumes the moment you stop. Say it, and it happens.
- Always-on tiny listener - about 0.3 GB, on the GPU or the CPU, runs beside your dictation engine
- 8 built-in commands, each a short 2-word phrase
- 2-word phrases chosen so the tiny model hears them reliably
- Auto-pauses during dictation, auto-resumes the second you stop
- Load or unload models by voice while idle ("warm up" / "cool down")
- The command listener is local too - your voice never leaves your PC
Vibe coders
Hands on the keys in Cursor or Copilot - with voice commands switched on in Settings, say "go ahead", dictate the next prompt, "wrap up", "copy that", paste. Ship without breaking flow to hunt for a hotkey.
Hands-busy & accessibility
Cooking, soldering, mid-workout, or resting your wrists from RSI - run dictation start to finish by voice, from across the room.
Streamers & presenters
Trigger capture and cleanup without alt-tabbing out of your scene. Your voice is the shortcut; the window can stay hidden.
Mini Mode - a pill that floats over everything.
Collapse Voxmelt into a tiny always-on-top pill you can drop anywhere and snap to any screen edge. It records through the same engine as the main window, mirrors your theme live, and shows the transcript and AI output in compact tabs - so you can dictate into any app without the full workbench in the way. Pair it with No-Hands Mode and the main window can stay closed in the tray while you run the whole flow from the pill. Out of the way, one tap away.
- Always-on-top, drag anywhere, snap-to-edge, optional lock
- Same engine as the main window - keeps working with the window closed to tray
- Live transcript + AI-output tabs in a 72px pill
- Full theme sync - 5 themes, custom colors, custom opacity
- One-tap copy, paste into any app; mic mirrors the main record button
- "Ready halo" lights up when No-Hands Mode is armed
Voice-input power users
Park the pill in a corner and dictate into Slack, email, chat, or docs all day - no full workbench hogging the screen.
Small screens & dual-monitor
Keep your work full-screen; the pill rides the edge of a laptop panel or a second display, always one tap away.
Vibe coders
Float it over the IDE and talk your prompts straight into Copilot or Cursor - eyes on the code, not on a separate app.
Your silicon is ready to melt.
NVIDIA Parakeet dictation on the CPU you already own - and if there is an NVIDIA GPU in your rig, a full local AI studio on top.
Direct installer, for PCs without the Store; Windows will warn because it is not code-signed.