Bulk Video and Audio Transcription Tool

Batch transcribe video and audio files offline into text, and export to TXT, SRT, VTT, JSON, LRC, CSV and other subtitle and text formats with one click.

Bug report

Loading tool, please wait...

If the online tool fails to load or run, try the desktop tool.https://tools.yikeaigc.com/

Tool Usage

Return to old version

A desktop version is available with more powerful, stable features. Open desktop version

Instructions

Software Usage Instructions

Add files to recognize: Click the "Select File" or "Select Folder" button to add video or audio files that need to be converted to text. You can also drag files or folders directly into the software window. Batch mixed adding of common video and audio formats is supported, including MP4, MKV, MOV, AVI, FLV, WebM, MP3, WAV, FLAC, M4A, and more.
Set the save directory: Choose where to save the processed subtitles and text. Drag-and-drop setting is supported. If "Traverse subdirectories" is checked, you can also check "Keep original path structure"; the output files will correspond one-to-one with the directory structure of the source folders.
Select the output format:
- TXT: Plain text, suitable for organizing manuscripts and meeting minutes
- SRT / VTT: Subtitle formats with timestamps, which can be directly attached to videos
- JSON: Structured data, convenient for secondary development and program processing
- LRC: Lyrics format, suitable for music and audiobooks
- CSV: Table format, convenient for editing and proofreading in Excel
Configure recognition parameters:
- Model path: Uses the Paraformer Chinese recognition model, FSMN voice activity detection, and CT punctuation model by default
- Batch processing duration: Set the maximum audio duration for a single inference
- Maximum VAD segment milliseconds: Control the granularity of speech segment splitting
- Hotword list: Enter proper nouns, names, and brand terms to improve recognition accuracy
- Skip/overwrite strategy: Choose whether to skip or overwrite existing output files
Start batch recognition: Click the "Start Recognition" button, and the software will automatically extract audio tracks, split speech segments, run inference transcription, and add punctuation. During processing, you can view progress in the log area. Click the "Stop" button if you need to cancel.
View and use results: After recognition is complete, go to the configured save directory to view the files in the corresponding formats. SRT/VTT subtitles can be dragged directly into editing software or media players.

FAQ

A: Video formats include MP4, MKV, MOV, AVI, FLV, WMV, WebM, M4V, MPG, TS, 3GP, RMVB, and more; audio formats include MP3, WAV, FLAC, AAC, OGG, M4A, WMA, OPUS, AC3, AMR, APE, and more. Video and audio files can be added together in the same batch, and the software will automatically recognize them and extract the audio tracks.

A: The software uses local offline recognition and does not require an internet connection at any stage. The model files are stored in the models folder in the program root directory. Based on FunASR's Paraformer Chinese model, it provides high recognition accuracy for Mandarin. Using the hotword list can further improve recognition of proper nouns, names, and terminology.

A: TXT is suitable for organizing manuscripts and writing materials; SRT and VTT are common subtitle formats that can be attached to videos and web players; JSON contains complete timestamps and segment information, making it convenient for secondary development; LRC is suitable for music and audiobooks; CSV makes it easy to review and edit by time segment in Excel. You can select multiple formats for export at the same time.

A: Enter proper nouns that are easily misrecognized in the hotwords input box, such as people's names, company names, product names, technical terms, etc. Separate multiple words with spaces or line breaks. The recognition engine will prioritize matching these words, with more noticeable results for vertical industry content, self-media voiceovers, corporate training, and similar scenarios.

A: This option becomes available after selecting "Traverse subdirectories." When enabled, output files will be generated in corresponding subdirectories according to the folder hierarchy of the source folder, making it easier to archive and manage large batches of materials. If duplicate file names are encountered, a number will be automatically appended to the file name to distinguish them.

A: The software has built-in VAD voice activity detection, which first automatically splits long audio by silent segments, then sends the segments to the recognition model according to the "batch processing duration" parameter. In theory, it can fully process long videos lasting several hours. Recognition speed is related to your computer's CPU performance. When processing large batches of materials, it is recommended to close other high-resource programs.

A: The default model is Chinese Paraformer, which provides good support for Mandarin and mixed Chinese-English content. Its recognition performance for dialects such as Cantonese, Sichuanese, and Hokkien is relatively average. If your material is mainly in a dialect, we recommend adjusting the pronunciation style first or choosing a better-matched model to replace the default path.

A: The software enables the CT-Punc punctuation model by default, which automatically adds punctuation such as commas, periods, and question marks to the recognition results. If there is no punctuation, please check whether the punctuation model path in the parameters correctly points to the ct-punc directory under the models folder, and make sure the model files are complete.