txt2mp3
Summary
Subscription TTS billing turns a podcast side project into a monthly invoice that scales against you — every draft, every revision, every re-read charged to a meter you didn't budget for. txt2mp3 is a macOS app that breaks that model with a one-time payment and a model that never phones home.
The core loop is paste-pick-generate: drop a script into the Studio screen, select one of fourteen preset voices across seven accents, and the app writes an MP3 to disk — 48 kHz, with character-level timestamps, zero bytes uploaded. Voice cloning works from a short reference recording that stays on the machine. A local MCP server lets Claude, Cursor, or any MCP-compatible agent call list_voices and speak without touching a cloud API. The ceiling appears fast: fourteen voices is a fixed library, and the app ships for Apple silicon only, so Windows and Linux teams are blocked before they start.
Bottom line: For a macOS creator or developer who needs unmetered offline TTS with a one-time license and local agent integration, this fits cleanly — but a team needing voice variety beyond fourteen presets, or running on any platform other than Apple silicon, will exhaust the library or the OS constraint before they exhaust the subscription math.
Pricing Plans
- Price
- $29 one-time (rises to $69 then $99)
- Free Tier
- Limited text length per generation in trial; all features otherwise available
License
One-time payment for unlimited generation and v1 updates on up to 5 Macs
- All voices
- Voice cloning and design
- MP3 export with timestamps
- Local MCP server
View full pricing on txt2mp3.app →
Pricing may have changed since last verified. Check the official site for current plans.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Pros
Sign in to edit- One-time license covering all v1 updates, so a project that generates thousands of takes — drafts, revisions, alternate reads — never accumulates a per-character bill.
- Voice cloning from a short reference recording that stays on-device, which means a creator can produce personalized narration without uploading someone's voice to a third-party server.
- Local MCP server binding to localhost, so an AI agent can generate and save speech files without routing audio through an external API or acquiring cloud credentials.
- 48 kHz MP3 export with character-level timestamps included, which means downstream tooling — subtitle sync, audio editing — has alignment data without a separate processing step.
- Commercial use rights ship with the license, so generated audio can go into a product, podcast, or client deliverable without a separate rights review.
Cons
Sign in to edit- The preset voice library is fixed at fourteen voices across seven accents. A project requiring more than fourteen distinct characters — or a voice the presets don't approximate — has no expansion path inside the app. Teams hit this wall at script complexity, not at scale, and move to ElevenLabs or a similar cloud provider that offers a broader model library.
- The app runs on Apple silicon only. A development team on Windows or Linux cannot run the app or its MCP server at all. There is no fallback, no web tier, no cross-platform CLI — the platform constraint eliminates the tool before any feature evaluation begins.
- The MCP server is a local process on a single machine, which means agent pipelines that need to generate audio from a remote server, a CI job, or a cloud-hosted agent cannot reach it. Teams needing remote or headless speech generation must maintain a separate cloud TTS integration alongside this one.
About
- Platforms
- macOS (Apple silicon)
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-09-20T14:35:26.143Z
Best For
Who it's for
- macOS users wanting offline TTS without subscriptions
- Developers needing local voice generation for agents or tools
- Creators requiring commercial rights to generated audio
What it does well
- Voice cloning from short recordings for personalized narration
- Local MP3 export for podcasts, audiobooks, or scripts
- Integration with AI agents via local MCP server for automated speech output
Integrations
Add notes, reviews, and benchmarks so the next visitor gets a clearer picture.
Compare txt2mp3
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeFrequently Asked Questions
- Is txt2mp3 free?
- txt2mp3 has a permanent free tier alongside paid upgrades (paid plans from $29 one-time (rises to $69 then $99)). You can keep using a baseline version indefinitely without paying.
- Is txt2mp3 open source?
- No — txt2mp3 is a closed-source tool. Source code is not publicly available.
- Does txt2mp3 have an API?
- Yes. txt2mp3 exposes a developer API. See the official documentation at https://txt2mp3.app for details.
- Can I self-host txt2mp3?
- Yes. txt2mp3 supports self-hosting on your own infrastructure.
- What platforms does txt2mp3 support?
- txt2mp3 is available on: macOS (Apple silicon).
Curated lists that include this category
Subscription TTS billing turns a podcast side project into a monthly invoice
The vendor states txt2mp3 runs entirely on-device after a one-time purchase. Users paste a script into the Studio screen, pick from fourteen preset voices across seven accents, and export a 48 kHz MP3 with character-level timestamps. No data leaves the machine.
Local cloning and agent access
Voice cloning starts from a short reference recording kept on the Mac. A built-in MCP server exposes list_voices and speak calls so Claude, Cursor, or other MCP clients can trigger speech output without cloud credentials or APIs.
Trade-offs
The preset library stays fixed at fourteen voices. The app ships only for Apple silicon, so Windows and Linux users have no path to run it or its MCP server. Trial mode limits text length per generation; full features otherwise work offline.
Who it is for / who should skip it
Best for macOS users who want offline TTS without subscriptions, developers integrating local voice output into agents, and creators who need commercial rights to the audio. Skip it if your team needs more than fourteen voices or runs on non-Apple hardware.