---

name: tts description: "Use Coqui TTS for released-model inference, the tts and tts-server commands, training and fine-tuning plans, vocoder/audio tooling, and FreeVC voice conversion workflows." disable-model-invocation: true metadata: disco-role: operating license: MPL 2.0

Coqui TTS

Use this skill for the Coqui TTS Python package and its installed console commands. It routes future agents to the right workflow without reopening the original repository.

Start here

1. Read references/package-overview.md for the package shape, public entry points, supported Python range, and verified runtime facts.
2. Run scripts/check_tts_environment.py to confirm the installed package, CLI help, registry access, and optional CUDA smoke in the current environment.
3. Read references/troubleshooting.md for cross-cutting install/import/version/audio/cache/backend failures.
4. If the repo snapshot may be stale, read references/repo-provenance.md before using or refreshing this skill.

Install and import check

• Supported Python: >=3.9, <3.12.
• Recommended install for a released environment: python -m pip install TTS==0.22.0.
• If you are actively maintaining a matching checkout, an editable install such as python -m pip install -e . is acceptable, but the runtime skill itself must not depend on the checkout remaining present.
• Minimal smoke check:

python -c "import TTS; from TTS.api import TTS as TTSApi; print(TTS.__version__); print(TTSApi)"

• Safer full smoke check:

python skills/disco/tts/scripts/check_tts_environment.py

Route map

• Python inference and model zoo: TTS.api.TTS, ModelManager, released-model loading, model-name grammar, custom checkpoint inference, tts_with_vc, XTTS/Fairseq/Bark/Tortoise naming cautions. Read sub-skills/inference-and-model-zoo/SKILL.md.
• Installed CLI and demo server: tts, tts-server, command construction, parser flags, pipe output, model metadata, and safe local server usage. Read sub-skills/server-and-cli/SKILL.md.
• Training, configuration, and data: dataset formatters, Coqpit configs, tokenizer/phonemizer setup, TTS training and fine-tuning plans, speaker embeddings, and recipe adaptation. Read sub-skills/training-config-data/SKILL.md.
• Vocoder and audio tools: vocoder configs/training, audio preprocessing, statistics, resampling, VAD trimming, and mel/audio compatibility. Read sub-skills/vocoder-and-audio-tools/SKILL.md.
• Voice conversion: FreeVC and TTS-with-VC workflows using source/target/reference wavs. Read sub-skills/voice-conversion/SKILL.md.

When to read this skill versus a sub-skill

• Start at the root when the user only says "Coqui TTS", "TTS", tts, tts-server, XTTS, FreeVC, "voice cloning", "vocoder", "training", or "dataset formatting" and you need to route the request.
• Go directly to a sub-skill when the workflow is already specific, for example "run the CLI", "use the Python API", "prepare a training config", or "convert a voice with FreeVC".
• Use the root only for cross-cutting install/import checks, package provenance, and choosing the correct route.

Cross-cutting notes

• Released models, model downloads, and some training workflows can touch network, cache, disk, or license/TOS prompts. Treat them as explicit user-acknowledged actions.
• CUDA is available on some hosts and can accelerate many workflows, but it is not required for the default package/import smoke checks in this skill.
• If a workflow needs audio-format repair, resampling, statistics, or VAD, route to the vocoder/audio sub-skill instead of duplicating that logic here.
• If a workflow needs speaker/language selection, custom checkpoint inference, or registry lookup, route to the inference/model-zoo sub-skill instead of rebuilding those rules here.
• If the repository snapshot changed, compare against references/repo-provenance.md before relying on this skill.