Transcribe audio and video
Turn speech into timed subtitles. Choose a local model and language; the app tests your device before the first run.
Transcribe, translate, and style subtitles—all on your device.
Transcribe audio or video into timed subtitles, then edit, translate, and style them alongside your media. Export SRT, VTT, or ASS files, or render a video with captions built in.
Turn speech into timed subtitles. Choose a local model and language; the app tests your device before the first run.
Play your media while editing subtitle text and timing. Fine-tune word boundaries where supported, with undo and redo.
Translate a full track or selected subtitles. Keep each language in its own track for review and editing.
Refine timing with alignment models and add speaker labels. Review the results before applying them.
Choose a preset or customize fonts, colors, and placement. Preview changes across a track or on one subtitle.
Control subtitle length, word count, and line breaks. Split at pauses and sentences to keep captions readable.
Save SRT, VTT, or styled ASS files. Export transcripts or JSON, including word and speaker details when available.
Save a video with styled captions built in. Your original media stays separate.
Reopen your media, tracks, styles, and saved runs. Projects stay in this browser, with local storage controls.
Create a project and choose an audio or video file.
Choose a model and language, then generate subtitles. The app downloads the model and tests your device on first use.
Check the text and timing, add translations or styling, then download subtitles or a captioned video.
We’ll email you when Local AI SRT / Subtitles Generator is ready.
Align a matching English transcript to audio locally in the browser. Returns complete word boundaries; Q4 WebGPU recommended, WASM CPU fallback measured independently.
Align an existing transcript to mono audio and estimate word/character start and end times using the official non-autoregressive 5000-class timestamp head. Does not transcribe unknown speech.
Align a matching transcript to its recording with the auxiliary CTC checkpoint distributed with NVIDIA Canary 2.
Translate English text into Spanish and return the translated text. This worker supports only English-to-Spanish translation, not automatic language detection or other language pairs.
Translate text or text visible in still images locally. Supply source and target language codes; review OCR, names, numbers and omissions.
Translate short text in four verified directions: English→Spanish, Spanish→English, English→French and Chinese→English. Tencent’s community license applies; review every translation.
Translate short multilingual text locally with the original HY-MT1.5 1.8B checkpoint and official source-aware translation prompt.
Translate text locally with Tencent Hy-MT2 1.8B using explicit language selections and the official translation prompt.
Local English to French translation using the original FP32 EuroNano Base checkpoint.
Run local text questions, still-image questions using Qwen3.5 4B Desktop. Answers can hallucinate or miss visual details.
Ask questions, draft short text, and experiment with multilingual conversations using a compact chat model. Returns an assistant reply; complex reasoning and factual answers need review.
Ask questions, draft short text, or continue a conversation with a compact multilingual chat model. Returns an assistant reply; answers can be inaccurate and should be checked.
Chat with text or images locally using Qwen3.5.
Chat with text, images or WAV audio using the complete Gemma 4 E2B and E4B checkpoints on verified configurations.
Transcribe speech with punctuation and approximate word-level timestamps. Native multilingual FastConformer/TDT decoding retains complete token timings and durations.
Transcribe authorized WAV files locally with the 1.5B streaming checkpoint. Native speaker-prefixed text and complete chunk outputs arrive progressively; hotword context is optional.
Transcribe WAV audio locally with word-level timestamps in seconds.
Transcribe speech locally in your browser with NVIDIA Parakeet TDT 0.6B v3. Returns punctuation, word and token timestamps, confidence scores and the complete native result.
English speech recognition with word-level CTC emission timestamps. Returns transcript, complete word timings and model confidence; no speaker labels or translation.
Multilingual transcription, English translation, and optional timestamps.
Multilingual transcription, English translation, and optional timestamps.
Transcribe local audio progressively with NVIDIA’s multilingual RNNT model. Returns complete text, approximate word-level timestamps and the native recognition result.
Estimate who spoke when locally using pyannote Community-1. Returns regular and exclusive speaker-turn timelines with anonymous speaker labels, not names or transcripts.
Identify anonymous speaker turns locally with pyannote segmentation and3D-Speaker embeddings. Returns a complete speaker timeline; WASM CPU only.
Identify up to four speaker channels in speech recordings with incremental timed segments.
Estimate anonymous speaker turns from complete WAV recordings with the native eight-channel Nemotron3 diarization model.