SubtitleEdit/subtitleedit · error · Exception

Paddle OCR failed: {paddleOcr.Error}

Error message

Paddle OCR failed: {paddleOcr.Error}

What it means

Thrown after PaddleOcr.OcrBatch completes: the engine set a non-empty paddleOcr.Error AND every group's Text is still empty, i.e. OCR produced no usable result and the engine self-reported a failure. It is only raised when nothing was recovered, so a partial success (some groups with text) does not throw. The exception bubbles to the StartOcr catch which logs and shows a MessageBox.

Source

Thrown at src/ui/Features/Video/VideoOcr/VideoOcrViewModel.cs:948

                if (group != null)
                {
                    group.Text = VideoOcrLineBuilder.CleanOcrResult(p.Text);
                    reportProgress();
                    addPreviewLine(group);
                }
            });

            // The frames are already image files on disk, so pass them by file name -
            // one batch, no per-image decode/encode, memory stays flat.
            var batch = ocrGroups
                .Select((g, i) => new PaddleOcrBatchInput { Index = i, SourceFileName = g.RepresentativeFileName })
                .ToList();

            var paddleOcr = new PaddleOcr();
            await paddleOcr.OcrBatch(engineType, batch, language, mode, progress, cancellationToken);
            if (!string.IsNullOrEmpty(paddleOcr.Error) && ocrGroups.All(p => string.IsNullOrEmpty(p.Text)))
            {
                throw new Exception("Paddle OCR failed: " + paddleOcr.Error);
            }
        }
        else if (engineType == OcrEngineType.Ollama)
        {
            using var ollamaOcr = new OllamaOcr(Se.Settings.Ocr.OllamaOcrTimeoutMinutes);
            await RunLlmOcr(ocrGroups, group => OcrWithBitmap(group, bitmap =>
                    ollamaOcr.Ocr(bitmap, OllamaUrl, OllamaModel, OllamaLanguage, cancellationToken)),
                () => ollamaOcr.Error, reportProgress, addPreviewLine, cancellationToken);
        }
        else if (engineType == OcrEngineType.Glm)
        {
            var glmOcr = new GlmOcr(GlmApiKey);
            await RunLlmOcr(ocrGroups, group =>
                    glmOcr.Ocr(group.RepresentativeFileName, GlmUrl, GlmModel, GlmLanguage, cancellationToken),
                () => glmOcr.Error, reportProgress, addPreviewLine, cancellationToken);
        }
        else if (engineType == OcrEngineType.LlamaCpp)
        {

View on GitHub (pinned to 17a9f07487)

Solutions

1. Read the embedded paddleOcr.Error text in the exception message — it carries the backend's own failure reason.
2. Re-run PaddleOcrInstallHelper.EnsureInstalled to (re)download the engine and its models.
3. Verify the selected Paddle language has its model files present and matches the installed backend version.
4. If using GPU, confirm the paddlepaddle-gpu build matches the CUDA version; fall back to CPU mode to isolate.

Example fix

// before
var paddleOcr = new PaddleOcr();
await paddleOcr.OcrBatch(engineType, batch, language, mode, progress, cancellationToken);
if (!string.IsNullOrEmpty(paddleOcr.Error) && ocrGroups.All(p => string.IsNullOrEmpty(p.Text)))
    throw new Exception("Paddle OCR failed: " + paddleOcr.Error);

// after - preflight the install and include which group count failed
if (!await PaddleOcrInstallHelper.EnsureInstalled(Window!, _windowService, engineType))
    throw new Exception("Paddle OCR engine is not installed.");
var paddleOcr = new PaddleOcr();
await paddleOcr.OcrBatch(engineType, batch, language, mode, progress, cancellationToken);
if (!string.IsNullOrEmpty(paddleOcr.Error) && ocrGroups.All(p => string.IsNullOrEmpty(p.Text)))
    throw new Exception($"Paddle OCR failed on {batch.Count} frame(s): " + paddleOcr.Error);

if (!await PaddleOcrInstallHelper.EnsureInstalled(Window!, _windowService, engineType))
    throw new Exception("Paddle OCR engine is not installed.");
if (SelectedPaddleLanguage is null) throw new Exception("No Paddle language selected.");

try { await paddleOcr.OcrBatch(...); }
catch (Exception ex) when (ex.Message.StartsWith("Paddle OCR failed"))
{
    SeLogger.Error(ex, "Paddle OCR batch failed: " + paddleOcr.Error);
    await MessageBox.Show(Window!, Se.Language.General.Error, ex.Message, MessageBoxButtons.OK, MessageBoxIcon.Error);
}

When it happens

Trigger: Paddle OCR backend errored on every frame: missing/incompatible PaddleOCR model files, wrong language code, GPU/CPU mismatch, the paddle CLI/python invocation failed, or an unsupported image format fed to the batch.

Common situations: PaddleOCR not fully installed (model dir missing); selected language has no downloaded model; inference backend version mismatch (paddlepaddle-gpu vs CPU); frames are empty/all-black so detection fails and the engine logs an error.

Related errors

• {engine.Error}
• Could not extract the current frame - see log for the ffmpeg
• No frames were extracted from the video - see log for the ff
• {Se.Language.Ocr.CrispEmbedNotDownloaded}
• {error}

AI-assisted analysis of SubtitleEdit/subtitleedit@17a9f07487 (2026-08-13). Data as JSON: /api/errors/a4e45a6bb42a64c3. Report an issue: GitHub.