ftts
Made by JeffOperate FrankenTTS ftts. Use when you make WAV/raw PCM, fix robot JSON, enroll voices, inspect FTTSQ, benchmark FTTS_INT8, test resident mode, develop, or release; excludes other TTS tools.
From Jeffrey's Skills.md: professionally crafted skills for any AI agent, installed and kept in sync by the jsm CLI. Subscription only, $20/month.
Works With
Quality and Utility come from canonical skill metadata. Community ratings appear separately as stars when subscriber reviews exist.
Subscribe to download
Public previews are available, but downloads, CLI sync, and premium content require an active subscription.
Sign In to Subscribe- Full skill content with deterministic versioning
- CLI sync + local search index
- Access to Jeffrey’s premium library
Save to your account
Saved skills live in your cloud list. Run jsm sync on each machine to pull them locally.
What's New with this Skill
1Updates, enhancements, and release notes for this skill.
New: ftts skill, operating FrankenTTS on evidence instead of memory
ftts turns text into WAV or raw PCM using int8-quantized kernels by default (FTTS_INT8=0 restores the f32 route), and this skill is built on one rule: probe the exact binary and evidence surface before prescribing a command. Checkout, release, Homebrew formula, installed binary, model bundle, and browser build can all disagree, so the skill ships a dated boundary card and a truth stack for resolving them in order.
- The one rule: probe the exact binary before prescribing commands; release 0.1.8, an installed binary at 0.1.7, and a README badge at 0.1.4 have all been true at once.
- A dated cold-start boundary card records each verified boundary next to the exact command that re-verifies it.
- Model bundle facts on hand: 8 pinned files, 1,999,608,000 bytes, tracked by an embedded manifest.
- Voice enrollment produces a raw 4,096-byte .spk artifact; it carries no consent metadata, so consent has to be established outside the artifact.
- Robot JSON schema v1: file events on stdout with errors possibly stderr-only, so parsers are told to read both streams.
- Resident mode exists to avoid repeated model load; performance comparisons must state the resident policy and compare matched warm and cold conditions.
- Performance claims are gated on a same-invocation incumbent harness, not remembered benchmarks.
The surfaces are simple: ftts say for synthesis, ftts enroll for voices, ftts robot health for model diagnostics, and robot JSON schema v1 for machine consumers. The tool ships fast, and version skew is a documented fact rather than a hypothetical: at the time the skill was cut, source and release stood at 0.1.8 while an installed Homebrew binary read 0.1.7 and the README badge was stale at 0.1.4. Remembered syntax is therefore a liability, and the truth stack orders what to trust: governing AGENTS.md files, then the exact installed binary, then current source and the performance ledger,