docling-project/docling · error · ImportError

whisper_s2t is not installed. Please install it via `pip ins

Error message

whisper_s2t is not installed. Please install it via `pip install 'whisper-s2t-reborn[pyav]>=1.7.1'`.

What it means

The whisper_s2t transcriber (_WhisperS2TModel) imports the optional `whisper_s2t` package (the whisper-s2t-reborn distribution with the pyav extra, pinned >=1.7.1) when enabled and re-raises ImportError with the exact pip spec otherwise. It is a constructor-time missing-optional-dependency error for the faster batched ASR engine.

Source

Thrown at docling/pipeline/asr_transcriber.py:535

    """Transcriber using WhisperS2T with CTranslate2 backend for high-speed inference."""

    def __init__(
        self,
        enabled: bool,
        artifacts_path: Path | None,
        accelerator_options: AcceleratorOptions,
        asr_options: InlineAsrWhisperS2TOptions,
    ):
        self.enabled = enabled

        _log.info(f"artifacts-path: {artifacts_path}")
        _log.info(f"accelerator_options: {accelerator_options}")

        if self.enabled:
            try:
                import whisper_s2t  # type: ignore
            except ImportError:
                raise ImportError(
                    "whisper_s2t is not installed. Please install it via "
                    "`pip install 'whisper-s2t-reborn[pyav]>=1.7.1'`."
                )

            self.whisper_s2t = whisper_s2t
            self.asr_options = asr_options

            raw_device = decide_device(
                accelerator_options.device,
                supported_devices=asr_options.supported_devices,
            )

            self.device, self.device_index = self._parse_device(raw_device)
            _log.info(
                f"Available device for WhisperS2T: {self.device} (index: {self.device_index})"
            )

            self.model_identifier = asr_options.repo_id

View on GitHub (pinned to 61d76f1ff3)

Solutions

1. Install exactly what the message asks: pip install 'whisper-s2t-reborn[pyav]>=1.7.1'
2. Or switch asr_options back to an engine already installed (standard whisper via the asr extra)

Example fix

# before
# InlineAsrWhisperS2TOptions selected, package missing

# after
pip install 'whisper-s2t-reborn[pyav]>=1.7.1'

def whisper_s2t_available() -> bool:
    try:
        import whisper_s2t  # noqa: F401
        return True
    except ImportError:
        return False

try:
    _WhisperS2TModel(...)
except ImportError as e:
    if 'whisper-s2t' in str(e):
        raise SystemExit("pip install 'whisper-s2t-reborn[pyav]>=1.7.1'") from e
    raise

When it happens

Trigger: Enabling InlineAsrWhisperS2TOptions in an environment where whisper-s2t-reborn[pyav] is not installed — note the plain `asr` extra installs openai-whisper/mlx-whisper, not necessarily this engine, so selecting s2t requires its own install.

Common situations: Opting into the faster whisper_s2t engine for bulk audio transcription; assuming `docling[asr]` covers every ASR engine; fresh environments after copying config that selects s2t.

Related errors

• whisper is not installed. Please install it via `pip install
• The 'mail-parser' package is required to process email files
• The 'beautifulsoup4' package is required to process HTML fil
• The 'pylatexenc' package is required to process LaTeX files.
• The 'marko' package is required to process Markdown files. I

AI-assisted analysis of docling-project/docling@61d76f1ff3 (2026-08-14). Data as JSON: /api/errors/53bc7d89f620f0d5. Report an issue: GitHub.