Run any Skill in Manus with one click

audiocpp-deferred-tts

Stars46

Forks6

UpdatedJuly 20, 2026 at 18:58

Run TTS model families audio.cpp supports but GuideAnts does not ship — Qwen3 CustomVoice builtin speakers, VibeVoice multi-speaker dialogue, MioTTS, VoxCPM2, PocketTTS, Vevo2 voice conversion — by downloading the model and spawning a private audiocpp_server in the sandbox. Use when the user names one of these models or asks for a TTS capability the loaded catalog model rejects.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

Elumenotion

Elumenotion/GuideAnts

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software DevelopersComputer and Mathematical Occupations·SOC 15-1252

File Explorer

7 files

SKILL.md

readonly

name	audiocpp-deferred-tts
description	Run TTS model families audio.cpp supports but GuideAnts does not ship — Qwen3 CustomVoice builtin speakers, VibeVoice multi-speaker dialogue, MioTTS, VoxCPM2, PocketTTS, Vevo2 voice conversion — by downloading the model and spawning a private audiocpp_server in the sandbox. Use when the user names one of these models or asks for a TTS capability the loaded catalog model rejects.
metadata	{"guideants":{"enabled":true,"display_order":35,"requires_toolsets":["sandbox"]}}

audio.cpp deferred TTS families (experimental)

The GuideAnts TTS wrapper is strictly catalog-gated, but the container ships the full audiocpp_server binary with every released loader — including families GuideAnts deferred for packaging/UI reasons, not engine reasons. The blockers are download/conversion chores a script can do by hand. Deliverables are WAV files in Output/; these models never appear in the GuideAnts voice picker or the live voice path.

Consent rule

Several of these families clone from reference audio (voice_ref). That is a supported use when the speaker consents (the user's own voice, or stated permission). Decline only imitation of a third party without consent.

Preflight

python3 Output/Skills/audiocpp-deferred-tts/scripts/preflight.py --for deferred-tts

Trust its verdict over this document. Heed its VRAM warning: a second engine competes with the wrapper's loaded models — prefer asking the user to unload the GuideAnts TTS model first (Settings, or curl -X POST http://127.0.0.1:8084/admin/unload). Ask, never do it silently.

The pattern

python3 Output/Skills/audiocpp-deferred-tts/scripts/fetch_model.py <hf-repo> --dest <dir> [--include <prefix>]
# any per-family conversion step — see references/deferred-models.md
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py start --path <dir> --family <family> --task tts
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py status   # poll until ready
python3 Output/Skills/audiocpp-deferred-tts/scripts/engine_tool.py speech "Hi" \
  --engine-url http://127.0.0.1:18099 --model <id> [--voice <speaker>] -o Output/hi.wav
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py stop     # always stop when done

Per-family recipes (repos, file scoping, conversions, known speaker ids) are in references/deferred-models.md — start with Qwen3 CustomVoice, it has the fewest new variables. Engine endpoints, config schema, and task tokens are in references/engine-api.md.

Ground rules

Script budget is ~5 minutes per call. spawn_engine.py start detaches and returns; poll status in a follow-up call instead of blocking. Big downloads resume across fetch_model.py calls.
Always stop the private engine when the task is done or the user moves on.
If the detached engine does not survive between script calls (preflight reports this), the flow is limited to synth requests issued in the same script call that spawned the engine — workable for small models only.
The engine errors loudly on unsupported options; surface that text.

Not possible with the container binary (don't burn time)

kokoro_tts and parakeet_tdt loaders are not in release builds, and some families are downloader-only upstream — no download fixes a missing loader. Those need the user's own custom build: see the audiocpp-host-tts skill.

Reporting

This skill exists to answer "is it possible?". End by telling the user which steps worked and what was blocked and why, quoting preflight/engine evidence, so the experiment log stays honest.

More from this repository

same repository

audiocpp-asr-extended

Elumenotion/GuideAnts

Transcription beyond the GuideAnts wrapper contract: transcribe workspace files by path (no upload, no 50 MB gateway cap), pass language hints, and sideload other qwen3-family ASR snapshots from Hugging Face through the wrapper. Use when a transcription needs a language hint, the file is large, or the user wants to try an ASR model the catalog doesn't list.

2026-07-2046

audiocpp-diarize

Elumenotion/GuideAnts

Speaker diarization: figure out who spoke when in a recording and produce a speaker-labeled transcript, by spawning a private audio.cpp engine with the sortformer diarization model and labeling each turn with the local ASR engine. Use when the user asks to diarize a meeting or call recording, separate or identify speakers, or wants a transcript broken down by speaker.

2026-07-2046

audiocpp-host-tts

Elumenotion/GuideAnts

Synthesize speech against the user's own host-native audiocpp_server build (outside the GuideAnts container) — including model families the container binary lacks, like Kokoro or Parakeet forks. TTS only. Use when the user says they run their own audio.cpp server or asks for a family the container build cannot load.

2026-07-2046

audiocpp-extended

Elumenotion/GuideAnts

Experimental: use audio.cpp scenarios and models that GuideAnts does not ship — speaker diarization (who spoke when, speaker-labeled transcripts), voice cloning from user-supplied audio, deterministic seeds, builtin speakers, language forcing, and deferred TTS families (Qwen3 CustomVoice, VibeVoice, MioTTS, VoxCPM2, PocketTTS) — by calling the raw audiocpp_server engines directly or spawning a private engine in the sandbox. Use when the user asks for a voice, model, or audio scenario the built-in audio tools reject or cannot express, including diarizing a meeting recording or labeling speakers in a transcript.

2026-07-2046

audiocpp-tts-controls

Elumenotion/GuideAnts

Advanced synthesis controls on the loaded audio.cpp TTS model: deterministic output via seed, forcing the spoken language, voice-design from a text description (instructions), and enumerating builtin speaker ids — none of which the built-in GuideAnts audio tools expose. Use when the user wants reproducible audio, a specific language, a described voice, or a list of available speakers.

2026-07-2046

audiocpp-voice-clone

Elumenotion/GuideAnts

Clone a voice from a user-supplied reference clip and synthesize speech with it, by calling the raw audio.cpp TTS engine with voice_ref/reference_text — fields the built-in GuideAnts audio tools do not expose. Use when the user wants generated speech to sound like their own voice or another consenting speaker's recording.

2026-07-2046

name	audiocpp-deferred-tts
description	Run TTS model families audio.cpp supports but GuideAnts does not ship — Qwen3 CustomVoice builtin speakers, VibeVoice multi-speaker dialogue, MioTTS, VoxCPM2, PocketTTS, Vevo2 voice conversion — by downloading the model and spawning a private audiocpp_server in the sandbox. Use when the user names one of these models or asks for a TTS capability the loaded catalog model rejects.
metadata	{"guideants":{"enabled":true,"display_order":35,"requires_toolsets":["sandbox"]}}

audio.cpp deferred TTS families (experimental)

Consent rule

Preflight

python3 Output/Skills/audiocpp-deferred-tts/scripts/preflight.py --for deferred-tts

The pattern

python3 Output/Skills/audiocpp-deferred-tts/scripts/fetch_model.py <hf-repo> --dest <dir> [--include <prefix>]
# any per-family conversion step — see references/deferred-models.md
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py start --path <dir> --family <family> --task tts
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py status   # poll until ready
python3 Output/Skills/audiocpp-deferred-tts/scripts/engine_tool.py speech "Hi" \
  --engine-url http://127.0.0.1:18099 --model <id> [--voice <speaker>] -o Output/hi.wav
python3 Output/Skills/audiocpp-deferred-tts/scripts/spawn_engine.py stop     # always stop when done

Ground rules

Script budget is ~5 minutes per call. spawn_engine.py start detaches and returns; poll status in a follow-up call instead of blocking. Big downloads resume across fetch_model.py calls.
Always stop the private engine when the task is done or the user moves on.
If the detached engine does not survive between script calls (preflight reports this), the flow is limited to synth requests issued in the same script call that spawned the engine — workable for small models only.
The engine errors loudly on unsupported options; surface that text.

Not possible with the container binary (don't burn time)

Reporting

This skill exists to answer "is it possible?". End by telling the user which steps worked and what was blocked and why, quoting preflight/engine evidence, so the experiment log stays honest.