Перейти к содержанию

AudioFoundationSpeechFoundation

ConversationTTS

Assigns an explicit conversation speaker and caps the generated audio duration.

Text to speechVoiceHub-nativeconversationttsParameters: Not reportedLanguages: en, zh +1Training: nativeLicense: CC-BY-NC-4.0

Parameter metadata: Not reported: the audited metadata available for the registered default does not provide an exact parameter total.

Usage

Complete the VoiceHub installation once, then run this repository-authored example. Model pages intentionally contain no package-install command.

This example is maintained against VoiceHub's public API; it is not copied from an upstream demo or package README.

Model-specific path: Assigns an explicit conversation speaker and caps the generated audio duration.

Inputs and controls: Use stable integer speaker IDs when building multi-turn context.

from pathlib import Path

from voicehub import AutoModelForTextToSpeech, TTSGenerationConfig

model = AutoModelForTextToSpeech.from_pretrained(
    'AudioFoundation/SpeechFoundation',
    model_type='conversationtts',
    device="cuda",
    lazy_load=True,
)
output = model.generate(
    'VoiceHub keeps model integrations explicit and reproducible.',
    generation_config=TTSGenerationConfig(
        seed=42,
        output_file=Path("output.wav"),
    ),
    speaker=0,
    max_audio_length_ms=30_000,
    temperature=0.9,
    top_k=50,
)
print(output.file_path, output.sample_rate, output.metadata)

Use authorized recordings. Verify hardware needs and pin a revision in production.

Overview

conversationtts is a VoiceHub text to speech integration. This page is generated from its registry contract. Open the conversationtts Colab notebook.

Property Value
Task Text to speech
Architecture conversationtts
Runtime VoiceHub-native
Languages en, zh, yue
Capabilities text-to-speech, voice-cloning, conversation, multilingual, fine-tuning, safetensors, voicehub-native, native-runtime, raw-audio-fine-tuning, preencoded-code-fine-tuning, noncommercial
Reusable components —
Normalized output TTSOutput

Language support

Supported language abbreviations

en, zh, yue

These are the languages explicitly named in the upstream release README's podcast data.

Paper and GitHub

Configuration

Load configuration without constructing the model:

from voicehub import AutoConfig

config = AutoConfig.for_model('conversationtts')
print(config.model_type)
Property Value
Canonical model type conversationtts
Configuration class ConversationTTSConfig
Architecture class ConversationTTSForTextToSpeech

Processing

Create the registered processor without allocating model weights:

from voicehub import AutoProcessor

processor = AutoProcessor.from_pretrained(
    'AudioFoundation/SpeechFoundation',
    model_type='conversationtts',
)
print(type(processor).__name__)

Inference

The Usage example returns TTSOutput through AutoModelForTextToSpeech.

Input and output contract

Property Value
Readiness integrated-raw
Data architecture codec-lm
Sample rate 24,000 Hz
Contract getter get_tts_dataset_spec('conversationtts')
Variant Required fields One of Boundary Other rules
raw-text-audio — text / texts; audio / audio_values Source at most one: text / texts; audio / audio_values; forbidden: text_token_ids, text_ids, audio_codes, codes
raw-text-code — text / texts; audio_codes / codes Prepared at most one: text / texts; audio_codes / codes; forbidden: text_token_ids, text_ids, audio, audio_values
tokenized-text-audio — text_token_ids / text_ids; audio / audio_values Prepared at most one: text_token_ids / text_ids; audio / audio_values; forbidden: text, texts, audio_codes, codes
tokenized-text-code — text_token_ids / text_ids; audio_codes / codes Prepared at most one: text_token_ids / text_ids; audio_codes / codes; forbidden: text, texts, audio, audio_values
multi-codebook-batch tokens, labels, tokens_mask — Prepared —

Autoregressive text/audio-token or codec-language-model data. See the data workflow.

Training and optimization

Use available_optimization_passes() to discover reversible public passes. Unsupported runtime or hardware fails closed before mutation.

Training contract

Property Value
Support native
Family causal-lm
Recipe single-phase
Default phase codec_language_model
Training checkpoint AudioFoundation/SpeechFoundation
Native training graph yes
Phase Kind Components Required inputs Loss keys
codec_language_model objective model tokens, labels, tokens_mask loss, codebook0_loss, residual_loss

The integration accepts its declared source or prepared contract directly. Call model.validate_training_support() first, then follow the training workflow.

Checkpoints, provenance, license, and limitations

Property Value
Default checkpoint AudioFoundation/SpeechFoundation
Hugging Face ID AudioFoundation/SpeechFoundation
Repository availability verified through the Hugging Face model API on 2026-08-11; pin a revision before production use.
Checkpoint status Registry default; pin an immutable revision for production and reproducible evidence
Optional dependency extra Core package
Hardware and runtime Usage selects cuda; verify checkpoint-specific requirements
Real-checkpoint evidence Release evidence; a registry default alone is not execution evidence
Implementation voicehub.models.conversationtts.modeling_conversationtts.ConversationTTSForTextToSpeech
Configuration voicehub.models.conversationtts.configuration_conversationtts.ConversationTTSConfig
Source provenance voicehub/models/conversationtts/source/SOURCE.json
License CC-BY-NC-4.0

Source, checkpoints, datasets, and evaluation tools are non-commercial. Commercial use: not allowed.

Confirm the checkpoint revision, access terms, provenance, and license.

Limitations

  • No integration-specific checkpoint limitation is registered. Verify the selected checkpoint revision and its documented runtime requirements.
  • Validate memory, precision, and optional dependencies on the target system.
  • Public optimizations fail closed when the runtime or hardware cannot satisfy their validation contract; an unavailable pass is not reported as applied.
  • Contract tests do not replace the linked released-checkpoint evidence.

Public API

Use the stable configuration, processor, and task-model facades below.

Configuration

ConversationTTSConfig

View source

ConversationTTSConfig(**config_kwargs)

Parameters

  • **config_kwargs — Configuration fields validated by ConversationTTSConfig.

Model

ConversationTTSForTextToSpeech

View source

AutoModelForTextToSpeech.from_pretrained(
    pretrained_model_name_or_path,
    *,
    model_type='conversationtts',
    config=None,
    **model_kwargs,
)

Parameters

  • pretrained_model_name_or_path — Hub ID or compatible local directory.
  • model_type — Canonical model type; use 'conversationtts'.
  • config — Optional preloaded ConversationTTSConfig instance.
  • **model_kwargs — Model-specific loading arguments.
from voicehub import get_model_spec

spec = get_model_spec('conversationtts')
print(spec.display_name, spec.task.value)
Purpose Public object
Discover get_model_spec('conversationtts')
Load and run AutoModelForTextToSpeech
Configure ConversationTTSConfig
Process AutoProcessor
Model implementation ConversationTTSForTextToSpeech
Normalized output TTSOutput
Training contract get_training_spec('conversationtts')
Optimization lifecycle available_optimization_passes, apply_optimization_plan, optimization_manifest, restore_optimization_plan

See all model guides, inference, and the training matrix.