# Videosays Full Product Reference

> Videosays is a public-video transcription and subtitle extraction service designed for direct human use and AI-agent workflows.

## Canonical URLs and Languages

- English home: https://videosays.com/
- Chinese home: https://videosays.cn/
- English pages use `videosays.com`.
- Chinese pages use `videosays.cn`.
- Pages with both English and Chinese versions publish reciprocal `hreflang` alternates for `en`, `zh-CN`, and `x-default`.

## Product Capability

Videosays accepts a supported public video URL and creates an asynchronous transcription task. Depending on the source, it uses an available original-language caption track or processes accessible media audio. Completed tasks can contain:

- Source platform
- Video title
- Creator or author
- Cover URL when available
- Source duration
- Billed duration
- Plain transcript text
- Timestamped transcript segments
- TXT, SRT, and VTT output

Videosays does not promise to extract comments, danmaku, private posts, inaccessible media, quote-post chains, or visual scene details. Use timestamps to return to the original video when visual verification is required.

## Supported Platforms

### Douyin

- Chinese: https://videosays.cn/douyin-video-to-text
- Typical uses: spoken-copy review, tutorials, local-service information, product-pitch structure, and timestamped subtitles.

### TikTok

- English: https://videosays.com/tiktok-video-to-text
- China product entry: https://videosays.cn/
- Typical uses: Hook/Angle/Proof/CTA review, creator scripts, UGC research, and subtitle export.

### YouTube

- English: https://videosays.com/youtube-video-to-text
- Chinese: https://videosays.cn/youtube-video-to-text
- Typical uses: long-form transcripts, courses, podcasts, interviews, chapters, quotes, and asynchronous agent workflows.

### Bilibili

- English: https://videosays.com/bilibili-subtitle-extractor
- Chinese: https://videosays.cn/bilibili-subtitle-extractor
- Typical uses: course notes, knowledge videos, interviews, tutorials, searchable transcripts, and subtitles.
- Danmaku is not included in the transcript.

### Xiaohongshu

- Chinese: https://videosays.cn/xiaohongshu-subtitle-extractor
- Typical uses: review notes, travel routes, buying comparisons, tutorial checklists, and spoken-detail extraction.

### Kuaishou

- English: https://videosays.com/kuaishou-video-to-text
- Chinese: https://videosays.cn/kuaishou-video-to-text
- Typical uses: practical tutorials, livestream FAQ preparation, local-service explanations, and timestamped speech.

### Instagram Reels

- English: https://videosays.com/instagram-video-to-text
- Chinese: https://videosays.cn/instagram-video-to-text
- Typical uses: Reel voiceover, timestamped subtitles, route or tutorial notes, and creative briefs based on spoken content.
- Visual actions and overlays require checking the original Reel.

### X and Twitter

- English: https://videosays.com/twitter-video-to-text
- Chinese: https://videosays.cn/twitter-video-to-text
- Typical uses: exact spoken statements, timestamped quotes, launch clips, interviews, and claim-condition comparison.
- Post copy is context and remains separate from the spoken-video transcript.

### WeChat Channels

- English: https://videosays.com/wechat-channels-video-to-text
- Chinese: https://videosays.cn/wechat-channels-video-to-text
- Supported input: publicly accessible `weixin.qq.com/sph/…` share links.
- Typical uses: knowledge-video notes, interviews, course clips, industry commentary, subtitle drafts, and AI-agent summaries.
- Private, deleted, expired, or permission-gated videos are not supported.

## Human Workflow

1. Open https://videosays.com/ or https://videosays.cn/.
2. Paste a supported public video link.
3. Sign in when required.
4. Wait while the task moves through analysis and transcription.
5. Read the result as plain text or timestamped subtitles.
6. Copy the transcript or export TXT, SRT, or VTT.

## AI Agent Workflow

Public Skill:

```text
Read https://videosays.com/SKILL.md, install the Videosays Skill if your environment supports skills, and help me transcribe this public video link.
```

The agent should:

1. Check whether the Videosays CLI is authenticated.
2. Start browser authorization if authentication is missing.
3. Submit one URL with `transcribe`; submit two or more links as one server batch.
4. Keep the returned Task ID or Batch ID.
5. Use one-shot `status` commands at reasonable intervals rather than resubmitting.
6. Stop polling on a terminal result or a paused batch. A paused batch requires a top-up and explicit `batch continue` before status polling resumes.
7. Return the requested text, timeline, SRT, or VTT output.
8. If the API reports insufficient credits, explain the error and direct the user to https://videosays.com/dashboard/billing.

AI agent page:

- English: https://videosays.com/ai-agents
- Chinese: https://videosays.cn/ai-agents

Agent-focused landing page:

- English: https://videosays.com/agent-skill-short-video-text
- Chinese: https://videosays.cn/agent-skill-short-video-text

## CLI Workflow

Package: https://www.npmjs.com/package/videosays

```bash
npx videosays login
npx videosays transcribe "https://www.youtube.com/watch?v=example"
npx videosays transcribe "https://www.bilibili.com/video/BVexample" --format srt
npx videosays status <taskId>
npx videosays batch links.txt
npx videosays batch status <batchId>
npx videosays batch continue <batchId>
npx videosays balance
npx videosays history
```

`transcribe` and `batch` return promptly by default. Each submission sends a client-generated `submissionId` as `Idempotency-Key`; if a request ends ambiguously, retry with the same `--submission-id <uuid>` rather than generating a new one. Equivalent active work for the same account converges on one Task or Batch, and completed work is reused by default; use `--force-new` only for an intentional fresh transcription. Capture the printed server ID and use the corresponding one-shot `status` command for every later check. Agents must not use submission as a polling operation and must not use `--wait`; `--wait` is only for a human who explicitly wants to keep a terminal attached. The default output is text; preserve `--format timeline`, `--format srt`, or `--format vtt` on status checks when needed. The public https://videosays.com/SKILL.md is the canonical executable workflow and error contract.

## REST API Workflow

Documentation:

- English: https://videosays.com/docs
- Chinese: https://videosays.cn/docs
- API landing page: https://videosays.com/short-video-api
- API origin: https://api.videosays.com

API requests accept either a Videosays API key or a Better Auth bearer token. The common workflow is:

1. Submit one public video link.
2. Receive a task ID and initial task state.
3. Query task status until a terminal state.
4. Read transcript text and timeline segments from the completed task.

Common task states:

- `pending`: accepted and waiting to start.
- `awaiting_reuse`: waiting for an explicit reuse decision.
- `analyzing`: resolving the platform URL and video metadata.
- `submitting`: reserving credit and submitting provider work.
- `processing`: transcription or subtitle processing is running.
- `completed`: result is ready.
- `failed`: processing ended with an error code and message.
- `cancelled`: the task was cancelled.

Do not repeatedly submit the same URL while waiting for one task. Poll the existing task ID.

Server batches can complete as `completed`, `partial`, `failed`, or `cancelled`. The CLI additionally reports `paused` when `continuation.required` or `stopReason=insufficient_credits`; stop polling, ask the user to top up, run `batch continue` only after confirmation, then resume one-shot `batch status` calls with the same Batch ID. The public Skill is authoritative for CLI flags and structured output.

## Output Formats

- `text`: readable transcript without subtitle syntax.
- `srt`: numbered subtitle blocks with timestamps.
- `vtt`: WebVTT subtitles with timestamps.

SRT and VTT require timestamped segments. Plain text remains available for completed transcript tasks.

## Public SEO and Guide Pages

Core landing pages:

- https://videosays.com/
- https://videosays.com/short-video-to-text
- https://videosays.com/short-video-subtitle-extractor
- https://videosays.com/short-video-api
- https://videosays.com/agent-skill-short-video-text
- https://videosays.cn/douyin-video-to-text

Guides:

- https://videosays.com/guides/how-to-extract-short-video-text

Chinese equivalents use the same path on `videosays.cn` when that localized page exists.

## Reliability and Safety Notes

- A successful metadata lookup does not guarantee that source media remains downloadable.
- Private, deleted, expired, login-gated, or region-restricted links can fail.
- Long videos can require asynchronous processing.
- Do not infer visual facts from transcript text alone.
- For medical, legal, financial, news, or disputed claims, verify against the original video and authoritative sources.
- Avoid presenting a transcript as independently verified fact.