Skip to content

magpie_tts.md

Latest commit

 

History

History
85 lines (69 loc) · 3.81 KB

magpie_tts.md

File metadata and controls

85 lines (69 loc) · 3.81 KB
Raw

MagpieTTS

MagpieTTS Multilingual 357M is wired as --family magpie_tts for offline text to speech. It uses baked speaker context prompts, a decoder transformer, an autoregressive local transformer, and NanoCodec waveform decoding.

Install

python3 tools/model_manager_v2.py install magpie_tts_orig

The default package installs a standalone GGUF directory:

models/MagpieTTS-Multilingual-357M-GGUF

Quick Start

audiocpp_cli --task tts --family magpie_tts \
  --model models/MagpieTTS-Multilingual-357M-GGUF \
  --backend cuda \
  --language en \
  --text "The production coordinator reviewed the overnight audio report and sent one clear update." \
  --request-option voice_id=Sofia \
  --out out.wav

German example:

audiocpp_cli --task tts --family magpie_tts \
  --model models/MagpieTTS-Multilingual-357M-GGUF \
  --backend cuda \
  --language de \
  --text "Das Team überprüfte am Morgen die Aufzeichnungen und bereitete einen kurzen Bericht vor." \
  --request-option voice_id=Jason \
  --out out_de.wav

Model

Field Value
Family magpie_tts
Task tts
Modes offline
Model directory models/MagpieTTS-Multilingual-357M-GGUF
Languages ar-AE, ar-MSA, ar-SA, de, en, es, fr, hi, it, ko, pt-BR, vi, zh
Voice input Baked voice prompt through voice_id
Reference cloning Not exposed by this integration

Language note: the upstream NVIDIA checkpoint advertises 12 languages, including Japanese (ja). The current audio.cpp integration exposes only the language frontends listed above. The standalone GGUF embeds the model config and sidecar resources, but Japanese requires NeMo's JapanesePhonemeTokenizer / JapaneseKatakanaAccentG2p path backed by OpenJTalk-style text processing. That tokenizer path has not been ported to the native C++ frontend yet, so ja is intentionally not listed as supported here.

voice_id accepts either a baked speaker name from the package speaker map or a numeric speaker index.

Options

Option Values Default Meaning
--language <code> / --request-option language=<code> language code en Target synthesis language used by the Magpie text frontend.
--request-option voice_id=<name-or-index> packaged speaker name or index 0 Baked speaker context prompt.
--temperature / --request-option temperature=<f> float >= 0 0.6 Audio-token sampling temperature.
--top-k / --request-option top_k=<n> integer > 0 80 Audio-token top-k sampling limit.
--guidance-scale / --request-option guidance_scale=<f> float >= 0 2.5 Classifier-free guidance scale for decoder logits.
--max-tokens / --request-option max_tokens=<n> integer > 0 500 Maximum decoder frames per text segment.
--text-chunk-size / --request-option text_chunk_size=<n> integer > 0 300 Long-form text chunk budget.
--text-chunk-mode / --request-option text_chunk_mode=<mode> default, tag_aware, japanese, endline default Framework text chunking mode.
--seed / --request-option seed=<n> integer >= 0 0 Audio-token sampling seed.
--session-option magpie_tts.graph_arena_mb=<mb> integer MiB 1024 Reusable graph arena size for Magpie stages.
--session-option magpie_tts.weight_context_mb=<mb> integer MiB 2048 Weight metadata arena size.
--session-option magpie_tts.weight_type=<type> native, f32, f16, bf16, q8_0 native Matmul weight storage type when supported.
--session-option magpie_tts.conv_weight_type=<type> native, f32, f16 native Convolution weight storage type when supported.

For GGUF packages, leave weight options at native unless you are deliberately testing a conversion or storage policy.