LLM-ready text
Return a single clean string for summaries, embeddings, classification and retrieval pipelines.
Convert a YouTube video to clean text through one API.
Request plain text for indexing and language models, or keep timestamped segments for editing and playback. The current Capslane scope is deliberately focused on public YouTube videos.
Video to text request
GET /v1/transcriptcurl 'https://capslane.com/v1/transcript?url=dQw4w9WgXcQ&mode=auto&text=true' \
-H 'x-api-key: YOUR_API_KEY'Choose the representation your workflow needs.
Text and timestamped segments are two views of the same canonical transcript, so changing the output does not require processing the video again.
Media-ready segments
Keep offsets and durations when the product must jump to a quote or align copy with playback.
Adjustable chunk size
Group text into chunks between 50 and 10,000 characters while preserving a stable API shape.
Language context
Every response reports the selected language and the caption languages observed on the source.
Automatic fallback
Use auto mode when text is required even if the video does not expose captions.
Request traceability
A request ID connects client errors to workspace logs without storing transcript content in technical logs.
Use text without losing source timing.
Index the full transcript
Use text=true for search, classification, analysis or language-model prompts.
Keep segments for playback
Use text=false for quote navigation, chapters, subtitles and clip-selection tools.
Switch without reprocessing
Both representations are built from one cached canonical transcript.
Questions developers ask.
Which video sources does Capslane support?
The current public API supports public YouTube videos. It does not claim support for arbitrary uploads or other social networks.
Can the response be one plain text string?
Yes. Set text=true to receive the transcript as one string instead of timestamped segments.
Can I use the result for search or language models?
Yes. Plain text works for indexing and language-model input, while segmented output preserves timing for media workflows.
Will the same video be processed again?
Capslane uses a shared canonical cache. A repeat request can return the existing transcript with cached=true.
Try the API on a real YouTube video.
Start free with no credit card and move to a paid plan when the integration grows.