FastDrop Video Classifier
Classify raw video footage into editor-ready roles (Hook, A-Roll, B-Roll, Screen Recording, VO, Blooper) with confidence scores, suggested filenames, and…
Что умеет
- Fastdrop.Video.Classify: Submit one video for classification into an editor role (Hook, A-Roll, B-Roll, Screen Recording, VO, Blooper), with confidence, explanation, suggested filename, and folder pat
- Fastdrop.Batch.Classify: Submit up to 50 videos for classification in one call. Credits for the whole batch are pre-authorized upfront and refunded per-video on failure. Prefer this over repeated fast
- Fastdrop.Job.Status: Check the status and results of a single video job. Free — costs no credits. This call waits for the job rather than sampling it: it holds open for up to 60 seconds and returns as
Какие данные видит
Нужен ли аккаунт
Не нужен: сервер работает без входа
Classify raw video footage into editor-ready roles (Hook, A-Roll, B-Roll, Screen Recording, VO, Blooper) with confidence scores, suggested filenames, and folder paths. Also runs technical diagnostics (codec, resolution, VFR, health score), extracts keyframe thumbnails, generates perceptual-hash fingerprints for duplicate detection, detects clip candidates, transcribes and translates audio in 50+ languages with SRT/TXT output, and searches a batch of videos by transcript using keyword, semantic, or hybrid matching. Batch up to 50 videos per call. Processing is asynchronous, but the status tools wait for the job rather than sampling it — one call usually returns a finished result, so agents don't spend turns polling. Readiness checks answer inline in the same call. Requires a FastDrop API key sent as an X-API-Key header; the free tier includes 100 credits and needs no card.
Список инструментов сервера (13)
Технические названия из tools/list. Нужны только разработчикам.
| fastdrop.video.classify | Submit one video for classification into an editor role (Hook, A-Roll, B-Roll, Screen Recording, VO, Blooper), with confidence, explanation, suggested filename, and folder path. Returns immediately with a job_id — processing is asynchronous. Wait poll_after_seconds, then call fastdrop.job.status until status is 'completed' or 'failed'. Typically finishes in 10-60s depending on tier and video length. This tool covers every capability. Prefer it over the single-purpose tools (fastdrop.video.thumbnails, fastdrop.video.diagnostics, and friends) — one call with several capabilities is cheaper and faster than several calls. Args: video_url: Public HTTP/HTTPS URL of the video to classify. capabilities: Capabilities to run. Options: classify, thumbnails, diagnostics, duplicates, clips, transcribe, full_pipeline. Default: ["classify", "thumbnails", "diagnostics"]. processing_tier: Quality tier. "basic" (heuristic, fastest, 2 credits), "enhanced" (heuristic + AI fallback, 4 credits), or "premium" (full AI, 6 credits). Use "basic" unless the user asks for higher accuracy. api_key: Deprecated. Send an X-API-Key header instead. Returns: job_id plus credits charged and a poll_after_seconds hint. |
| fastdrop.batch.classify | Submit up to 50 videos for classification in one call. Credits for the whole batch are pre-authorized upfront and refunded per-video on failure. Prefer this over repeated fastdrop.video.classify calls — it is one round trip instead of fifty. Returns immediately with a batch_id. Wait poll_after_seconds, then poll fastdrop.batch.status. A 50-video batch typically takes several minutes. Args: video_urls: Public HTTP/HTTPS URLs (max 50). capabilities: Capabilities to run on each video. Default: ["classify", "thumbnails", "diagnostics"]. Requesting "transcribe" here costs 3 credits per video rather than 8: batches return the transcript and translation but cannot generate SRT/TXT subtitle files. Use fastdrop.video.transcribe when you need the files. processing_tier: "basic", "enhanced", or "premium". api_key: Deprecated. Send an X-API-Key header instead. Returns: batch_id, per-video job_ids, total credits, and a poll_after_seconds hint. |
| fastdrop.job.status | Check the status and results of a single video job. Free — costs no credits. This call waits for the job rather than sampling it: it holds open for up to 60 seconds and returns as soon as the job finishes. Most jobs complete within one call, so call it once and read the status. Status values: queued, downloading, processing, completed, failed. If the response still carries poll_after_seconds, the wait timed out with the job unfinished — wait that many seconds and call again. If it has no poll_after_seconds, the job is terminal and calling again returns the same answer. Args: job_id: The UUID returned by a submission tool. response_format: "detailed" (default) returns everything including thumbnail URLs, diagnostics, and clips. "concise" returns only role, confidence, semantic label, and suggested filename/folder — use it when only the classification matters. api_key: Deprecated. Send an X-API-Key header instead. Returns: Job status, progress, and results when completed. |
| fastdrop.batch.status | Check the status and per-video results of a batch. Free — costs no credits. This call waits for the batch rather than sampling it: it holds open for up to 90 seconds and returns as soon as every video has finished. A large batch can outlast that; if the response still carries poll_after_seconds, wait that long and call again. No poll_after_seconds means the batch is done. Defaults to a concise projection because a 50-video detailed response contains 50 sets of presigned thumbnail URLs and can consume a large share of the context window. Ask for "detailed" only when you need thumbnails, clips, or diagnostics for every video — or better, call fastdrop.job.status on the one job you care about. Args: batch_id: The batch_id returned by fastdrop.batch.classify. response_format: "concise" (default) returns per-video status, role, confidence, and suggested filename. "detailed" returns the full payload including thumbnails, clips, and diagnostics. api_key: Deprecated. Send an X-API-Key header instead. Returns: Batch progress counts plus per-video results as they complete. |
| fastdrop.account.usage | Check the current credit balance and usage statistics. Free — costs no credits. Call this before submitting a large batch to confirm there are enough credits; a batch that exceeds the balance is rejected outright rather than partially run. Args: api_key: Deprecated. Send an X-API-Key header instead. Returns: Plan tier, credits remaining, and monthly usage broken down by capability. |
| fastdrop.batch.search | Search across every transcribed video in a batch by keyword or natural language. Returns timestamped hits showing where topics are discussed. Free — costs no credits. Works on any transcribed batch; classification is not required. The batch must have been submitted with the "transcribe" capability. Prefer several narrow queries over one broad one — results are more relevant and each response stays small. Long transcript snippets are truncated. Args: batch_id: The batch to search within. query: Search text (keyword or natural language). mode: "keyword" (full-text), "semantic" (embedding similarity), or "hybrid". max_results: Maximum results, 1-100. Default 20. Keep this low. api_key: Deprecated. Send an X-API-Key header instead. Returns: Hits with video_id, filename, start/end time, transcript text, relevance score, match type, and language. |
| fastdrop.capabilities.list | List every FastDrop capability with its credit cost, a description of what it does, and the available bundles. Free — costs no credits and needs no API key. Call this when you're unsure which capability name to pass to fastdrop.video.classify or fastdrop.batch.classify. If your client supports MCP resources, read 'fastdrop://capabilities' instead of spending a tool call. Returns: Capabilities (classify, thumbnails, diagnostics, duplicates, clips, transcribe) with per-video credit costs and descriptions, plus bundles. |
| fastdrop.video.readiness | Decide whether a video is fit for what you are about to do with it, before you spend anything doing it. Costs 1 credit. Answers in the same call rather than returning a job, because the point is to decide before committing. Use it ahead of an expensive pipeline — transcoding, tracking, a long transcription run — to avoid discovering halfway through that the footage was never going to work. The verdict is one of three: proceed nothing found that blocks this use fix_first something is wrong and can be fixed; a `fix` says how reject the footage cannot serve this purpose and no transformation helps — frames that were never captured cannot be added A `fix` is attached to every reason that has one, and to no reason that does not, so "is there a fix" is answerable without interpreting the text. Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. context: What the footage is for. See the field description. overrides: Per-call threshold changes. include_data: Raw measurements to echo back. Returns: A verdict with reasons, machine-actionable fixes, and a display-ready message. Very high-bitrate video is deferred instead: `job_id` is set in place of `verdict`, and fastdrop.job.status returns the same answer. |
| fastdrop.video.thumbnails | Extract 3-5 keyframe thumbnails from a video, returned as presigned URLs. Costs 1 credit. Shorthand for fastdrop.video.classify(capabilities=["thumbnails"]). Use that instead if you also want classification or diagnostics — one call is cheaper than two. Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. Returns: job_id plus a poll_after_seconds hint. Poll fastdrop.job.status for URLs. |
| fastdrop.video.diagnostics | Analyze a video's technical metadata: codec, resolution, FPS, variable-frame-rate detection, duration, container, and a 0-100 health score. Costs 1 credit. Shorthand for fastdrop.video.classify(capabilities=["diagnostics"]). Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. Returns: job_id plus a poll_after_seconds hint. Poll fastdrop.job.status for data. |
| fastdrop.video.fingerprint | Generate a perceptual hash fingerprint for duplicate detection. Costs 1 credit. Most useful across a batch — run fastdrop.batch.classify with the "duplicates" capability to fingerprint many videos and compare them. Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. Returns: job_id plus a poll_after_seconds hint. Poll fastdrop.job.status. |
| fastdrop.video.clips | Detect clip candidate timestamps with labels and confidence scores. Costs 2 credits. Shorthand for fastdrop.video.classify(capabilities=["clips"]). Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. Returns: job_id plus a poll_after_seconds hint. Poll fastdrop.job.status. |
| fastdrop.video.transcribe | Transcribe a video's audio, auto-detecting the language and optionally translating to English and generating subtitle files. Supports 50+ languages. Costs 5 credits — the most expensive single capability. This is the slowest capability; expect roughly a minute or more for long videos. To make a whole batch searchable, run fastdrop.batch.classify with the "transcribe" capability instead of calling this per video. Args: video_url: Public HTTP/HTTPS URL of the video. api_key: Deprecated. Send an X-API-Key header instead. language: Optional ISO 639-1 hint (e.g. 'en', 'es', 'ja'). Auto-detected if omitted — only set it when you already know the language. translate: Translate non-English transcriptions to English. Default true. output_formats: Generate subtitle files and return download URLs. Options: ["srt"], ["txt"], or ["srt", "txt"]. Omit for JSON only. Returns: job_id plus a poll_after_seconds hint. Poll fastdrop.job.status for the transcription, translation, segments, and any file URLs. |