naptha/tesseract.js · error

initialization failed

Error message

initialization failed

What it means

api.Init(null, langs, oem, configFile) returns -1 when Tesseract cannot initialize with the provided language data and engine mode. The library already attempts one automatic recovery: if cached data triggered the 'components are not present' debug message, it refreshes the cache and retries Init once. If the status is still -1 after that, it rejects with 'initialization failed'.

Source

Thrown at src/worker-script/index.js:305

        const debugStr = TessModule.FS.readFile('/debugDev.txt', { encoding: 'utf8', flags: 'a+' });
        if (dataFromCache && /components are not present/.test(debugStr)) {
          log('Data from cache missing requested OEM model. Attempting to refresh cache with new language data.');
          // In this case, language data is re-loaded
          await loadLanguage({ workerId, payload: { langs: loadLanguageLangsWorker, options: loadLanguageOptionsWorker } }); // eslint-disable-line max-len
          status = api.Init(null, langs, oem, configFile);
          if (status === -1) {
            log('Language data refresh failed.');
            const delCachePromise2 = langsArr.map((lang) => adapter.deleteCache(`${loadLanguageOptionsWorker.cachePath || '.'}/${lang}.traineddata`));
            await Promise.all(delCachePromise2);
          } else {
            log('Language data refresh successful.');
          }
        }
      }
    }

    if (status === -1) {
      res.reject('initialization failed');
    }

    res.progress({
      workerId, status: statusText, progress: 1,
    });
    res.resolve();
  } catch (err) {
    res.reject(err.toString());
  }
};

// Combines default output with user-specified options and
// counts (1) total output formats requested and (2) outputs that require OCR
const processOutput = (output) => {
  const workingOutput = JSON.parse(JSON.stringify(defaultOutput));

  const nonRecOutputs = ['imageColor', 'imageGrey', 'imageBinary', 'layoutBlocks', 'debug'];
  let recOutputCount = 0;

View on GitHub (pinned to a1ca80d9e3)

Solutions

1. Clear the cache (delete the cachePath directory or set cacheMethod: 'refresh' for one run) to force a clean re-download.
2. Match OEM to the data variant — legacy engine needs the full traineddata, LSTM needs the LSTM build; pass legacyCore/legacyLang consistently.
3. Verify the lang string format ('+'-separated ISO 639-3 codes) and that every code was actually loaded.
4. If passing a config object/string, simplify or remove it to isolate the cause.

Example fix

// before
const worker = await createWorker('eng', OEM.TESSERACT_ONLY); // LSTM data cached -> init -1

// after
const worker = await createWorker('eng', OEM.TESSERACT_ONLY, {
  legacyCore: true,
  legacyLang: true,
  cacheMethod: 'refresh', // force re-download of correct data
});

// Ensure OEM and data variant are consistent before init.
const OEM = require('tesseract.js').OEM;
function consistentOemData(oem, legacyLang, legacyCore) {
  const needsLegacy = oem === OEM.TESSERACT_ONLY || oem === OEM.TESSERACT_LSTM_COMBINED;
  if (needsLegacy && (!legacyLang || !legacyCore)) {
    throw new Error('Legacy OEM requested without legacyLang/legacyCore');
  }
}
consistentOemData(OEM.TESSERACT_ONLY, options.legacyLang, options.legacyCore);

const oemRequiresLegacyData = (oem) =>
  oem === 0 /* TESSERACT_ONLY */ || oem === 2 /* TESSERACT_LSTM_COMBINED */;

// On init failure, clear the cache once and retry with refreshed data.
async function initOrRefresh(langs, oem, opts) {
  try {
    return await createWorker(langs, oem, opts);
  } catch (e) {
    if (/initialization failed/i.test(e.message)) {
      return await createWorker(langs, oem, { ...opts, cacheMethod: 'refresh' });
    }
    throw e;
  }
}

When it happens

Trigger: Corrupted or truncated .traineddata; OEM mismatch such as requesting the legacy engine against LSTM-only language data; invalid/malformed lang string; partial download cached from a previous run; config file syntax error.

Common situations: Stale bad data in the cache directory from an interrupted download; using OEM.TESSERACT_ONLY with the default LSTM-only traineddata; custom config string that Tesseract rejects; disk/memory exhaustion during init.

Related errors

• Legacy model requested but code missing.
• `worker.detect` requires Legacy model, which was not loaded.
• Not found in cache
• Network error while fetching ${fetchUrl}. Response code: ${r
• Failed to load TesseractCore

AI-assisted analysis of naptha/tesseract.js@a1ca80d9e3 (2026-08-13). Data as JSON: /api/errors/01bdc160dc23edc1. Report an issue: GitHub.