Weftly

Cut one or more clips from any prior video job (find_clips, summarize, or video transcribe). Operates on a parent job — possessing the parent `source_job_id` is the capability, no upload step. Contrast with `clips_from_video`, which takes a local `filename` and uploads a video that has never been in Weftly before; use this tool whenever a video job already exists. Two choices are required on every new job, with no default — if the customer's intent is genuinely ambiguous, ask rather than guess: `orientation` (`horizontal` keeps the source framing; `vertical` crops to 9:16 for TikTok/Reels/Shorts) and `output` (`files` delivers one file per clip; `reel` concatenates every clip into a single file). Two-call flow: (1) call with `source_job_id` + `clips` (1+ objects `{start, end, title?}` in source seconds) + `orientation` + `output` to receive {job_id, payment_challenge}; (2) pay via MPP and call again with `job_id` + `payment_credential` to start processing. Price: `output: "reel"` bills a flat $0.50 regardless of segment count; `output: "files"` bills N × $0.50 for N clips, charged once for the whole batch (Stripe quantity). See /.well-known/mpp.json for the Tempo USDC rate. `output: "files"` caps a batch at 5 clips; `output: "reel"` concatenates every clip into one file, capped at 30 minutes of total output for `orientation: "horizontal"` or, for `orientation: "vertical"`, the selected profile's cap: 240s for `tiktok-primary` and `tiktok-primary-720p`, 180s for `instagram-reels`, 60s for `instagram-stories`. The 240s ceiling is an encode limit (crop+scale cost per segment against the container's own timeout), not a platform rule; the shorter profile caps are the destination's own limit. `orientation: "vertical"` accepts `profile` (default `tiktok-primary`; also `tiktok-primary-720p`, `instagram-reels`, `instagram-stories`) and `subject` (`center`/`auto`/`follow`, default `follow` — switches crop between active speakers across clip and segment boundaries, the mode that makes multi-speaker interview clips usable; `manual`/`subject_id` need per-clip data this tool cannot supply — use the REST API's `extract_vertical_clip` job type directly for those); both are rejected for `orientation: "horizontal"`, which keeps the source framing and has no crop to aim. `include_transcript` (default true) applies only to `output: "reel"` — mirrors extract_clip's clipped SRT + word-level transcript outputs — and is rejected for `output: "files"`. Outputs: `output: "files"` writes roles `clip-1-video` through `clip-N-video` plus a `clips-manifest` (JSON) recording each clip's timing, output role, and any per-clip failure — a batch delivering fewer than N clips is not refunded, only one delivering zero is; `output: "reel"` writes role `clip-video` (frame-accurate boundaries, 15ms audio fades at segment joins) plus, when `include_transcript` is true, roles `clip-srt` + `clip-words`. Every output is audio loudness-normalized to -14 LUFS / -1.5 dBTP; vertical output is additionally cropped to the selected profile's dimensions (1080×1920 by default) and burned with word-level captions by default (center-third placement, uppercase, one short line per card) — pass `captions: false` to render vertical output without them; horizontal output never gets captions. Payment: pay by credit card via the Stripe Checkout link (open the returned `payment_url` in any browser) or Tempo USDC via mppx; the challenge's WWW-Authenticate header and /.well-known/mpp.json are authoritative for which methods are offered. Source must still be in storage (72h TTL for find_clips parents, 24h elsewhere — check `expires_at` from get_job_status on the parent). Multiple clips_from_job calls against one parent are independent paid jobs. Retrying with `job_id` alone recovers the current state. Failed jobs auto-refund.

Find clips in a video AND cut up to 5 of them, under ONE payment. This is the alternative to calling find_clips and then clips_from_job yourself — prefer it whenever the user wants clips delivered end-to-end from a video they have not yet uploaded. CARD PAYMENT ONLY — this chain authorizes a Stripe hold (up to $4.50) before the final price is known (the total depends on how many clips are found and delivered), a shape MPP push settlement cannot express. A caller with no card-checkout capability (a wallet-only MPP agent) should NOT call this tool — call find_clips then clips_from_job instead, paying per step in USDC. Two-call flow: (1) call with `filename` and `content_type` (video only) — and an optional `query` to switch clip discovery from "best clips" to matching specific content (e.g. "the part about pricing") — to receive {job_id, status: "awaiting_payment", checkout_url}; (2) once the user says they have authorized payment, call again with the SAME `job_id` (no other fields needed) to recover state — this returns the same checkout_url if still unauthorized, or an upload step once the card hold is authorized (the client performs the upload, never you). Retrying with a NEW call instead of the same `job_id` authorizes a SECOND card hold for the same video — always resume with `job_id`, exactly like every other job-creating tool here. The card hold authorizes the worst case (one $2.00 find_clips analysis plus up to 5 × $0.50 clips), not a charge — the customer is charged once, at the end, for exactly what was delivered (fewer clips than the maximum is normal, not a failure). Clips are auto-selected by score and delivered as a `clips-manifest` output alongside the transcript and clip candidates on the same job. Poll with get_job_status(job_id) for the transcript and candidates; the clips themselves land as the job progresses through its chained stages. Set `orientation` to choose the shape: `vertical` (the default) cuts 9:16 clips for TikTok, Reels and Shorts; `horizontal` cuts 16:9 clips for YouTube or a website embed. `horizontal` CUTS the clips and hands them back — it does not upload or publish anything to YouTube; publishing is publish_to_youtube, a separate job the user has to ask for. Pick from what the user says they want the clips FOR, and if that is genuinely ambiguous, ask rather than guessing — the two produce different files at the same price. `orientation: "vertical"` clips are burned with word-level captions by default (center-third placement, uppercase, one short line per card); pass `captions: false` to render them without. `orientation: "horizontal"` clips never get captions.

Confirm that the file has been uploaded (via HTTP PUT to the upload_url from transcribe or summarize) and start processing. Verifies that the file is present in storage and that the job has been paid. Returns status "processing". Poll get_job_status to track progress and retrieve download URLs when done.

Cut and re-assemble a video from an edited copy of its own word-level transcript — the text-editing counterpart to clips_from_job's timestamp-based cutting. Operates on a parent job (find_clips, summarize, or video transcribe) that has a word-level transcript; possessing the parent `source_job_id` is the capability, no upload step. Get the editable text first via get_job_status(job_id, format: "edit") or the download route — one `[NNNN]`-marked line per sentence — then edit it in any text editor: delete words or whole lines, move whole lines, but do not add or change a word (a marker used twice, or any added/changed word, invalidates the whole edit and is rejected before payment). Two-call flow: (1) call with `source_job_id` + the full edited text as `edited_transcript` (sent inline, max 262144 UTF-8 bytes) + optional `title` to receive {job_id, payment_challenge} — an invalid edit is rejected here, before any charge, with the exact failing lines; (2) pay via MPP (Tempo USDC) or open the Stripe Checkout `payment_url` in a browser, then call again with `job_id` + `payment_credential` to start processing. Flat price: $1.00 — see /.well-known/mpp.json for the Tempo USDC rate. Up to 300 cuts and 60 minutes of output per call; repeated phrases match their first occurrence, left to right. Outputs: role `clip-video` (the edited video), plus `clip-srt` and `clip-words` re-timed to match the edit. Source must still be in storage (72h TTL for find_clips parents, 24h elsewhere — check `expires_at` from get_job_status on the parent). Use whenever the edit is easier to describe in text than in timestamps — tightening a clip or a short segment by deleting filler words and tangents, or reordering a handful of sentences; use `extract_clip` (via clips_from_job) instead when you already know the exact start/end seconds you want. Retrying with `job_id` alone recovers the current state. A rejected edit creates no job, so show its errors to the user from the response that returned them. Failed jobs auto-refund.

START HERE for any clip workflow on a video — `find_clips` is the canonical entry point and includes a full transcription as a free byproduct. **Do not call `transcribe` first**: doing so doubles the upload, doubles the spend, and produces the same transcript. Identify ranked candidate clips in a video — what to cut for highlights, social, or testimonials. Three-call flow: (1) call with `filename` (and optional `query`) to receive {job_id, payment_challenge}; (2) pay via MPP, then call with `job_id` + `payment_credential` to receive {upload_url} (presigned PUT, 1h expiry); (3) PUT the bytes, then complete_upload(job_id), then poll get_job_status(job_id). On completion, get_job_status returns three outputs: role `clip-candidates` (JSON matching /.well-known/weftly-clips-v1.schema.json — includes `source_job_id` and `source_expires_at`), role `transcript` (SRT, free byproduct), role `transcript-words` (JSON matching /.well-known/weftly-transcript-v2.schema.json, free byproduct). Each candidate carries `transcript_text` — the full text of what's in the clip — so callers can preview content before paying for clips_from_job. Optional `query` parameter switches to query mode (e.g., "they discuss pricing", "the part about hiring") with the same output shape; the `mode` field in clip-candidates.json indicates which mode produced the result. Flat price: $2.00 video — see /.well-known/mpp.json. **Source-reuse contract:** the source video stays in storage for 72h after find_clips completes. Hand the find_clips `job_id` (also returned as `source_job_id` in the candidates JSON) to `clips_from_job` as its `source_job_id` — within those 72h it cuts directly from the stored source: no re-upload, no re-transcribe, just $0.50 per cut (more for a multi-clip `output: "files"` batch). Pass the same `source_job_id` to as many `clips_from_job` calls as you need. Use for interviews, podcasts, sales calls, all-hands recordings. Retrying with `job_id` alone returns current state. Failed jobs auto-refund.

Check the status of a transcribe or summarize job. Returns the current state and, when completed, an `outputs` array. Each output has either `content` (returned inline) or a presigned, time-limited (1 hour) `download_url`. Small text outputs (e.g. `transcript` SRT, `clip-candidates`, `summary`) come inline as `content`; larger outputs — `transcript-words` JSON for any non-trivial recording, plus video outputs like `clip-video` / `clip-vertical-video` — come as a `download_url` to fetch when needed. Optionally pass `format` (srt, txt, vtt, json, words, edit) to get the transcript content inline in the top-level `transcript` field — `txt` and `vtt` are derived from the stored SRT; `json` is v1 (segments only); `words` is v2 (segments + per-word timestamps matching /.well-known/weftly-transcript-v2.schema.json); `edit` is a plain-text "[NNNN] words..." paragraph list, one line per paragraph, for editing by hand — send the edited text to edit_from_transcript to cut a re-timed video from it. Poll this periodically after calling complete_upload — wait at least 60 seconds between checks. For files under 10 minutes, jobs usually complete within 1-2 minutes. For long files (1hr+), expect 10-30 minutes. Derivative jobs (`extract_clip`, `extract_vertical_clip`, `publish_youtube`, `clips_vertical`, `clips_horizontal`, `edit_from_transcript`) have no upload step and no `awaiting_upload`/`awaiting_complete_upload` states — a paid derivative job reports `processing` once its parent source is confirmed still in storage, or a terminal `source_expired` if the parent source has expired: stop polling, and re-run the parent job named in `source_job_id` to get a fresh source before retrying. `publish_youtube` is the one exception: its upload starts by itself as soon as payment settles, so a paid job goes straight to `publishing`. If it reports `awaiting_trigger` instead, the upload did not start — almost always because the YouTube account is not connected yet; connect it and the publish can be resumed. A completed `publish_youtube` job's `outputs` array carries a `youtube_url`-roled entry with a `url` field pointing straight at the published video, instead of a `download_url`. Also use this to recover from lost state: if the original challenge was lost, call get_job_status(job_id) to retrieve a fresh challenge (status "awaiting_payment") or the upload URL (status "awaiting_upload").

Smoke-test the MPP payment plumbing end-to-end via this MCP server, for $0.01 USDC. Two-call flow: (1) call with no arguments to receive an MPP `payment_challenge`; (2) pay via MPP and call again with `payment_credential` set to the resulting Authorization header value (e.g. "Payment eyJ...") to receive {paid: true, timestamp, receipt_ref, payment_method}. Uses the exact same `createPayToAddress` + `createMppHandler` verification path as paid product tools (transcribe, summarize), so a green run here means real paid calls will work too. Stateless — no job is created, no database row written. Use this whenever you want to confirm a wallet, the MCP transport, the worker, and the production payment middleware are all healthy without paying a transcribe price. Cost: $0.01 USDC per attempt.

Summarize an audio or video file — returns both a text summary AND the full transcript (with per-word timestamps). Do not also call transcribe on the same file. Three-call flow: (1) call with `filename` to receive {job_id, payment_challenge}; (2) pay via MPP, then call with `job_id` + `payment_credential` to receive {upload_url} (presigned PUT, 1h expiry); (3) PUT the bytes, then complete_upload(job_id), then poll get_job_status(job_id). On completion, get_job_status returns three outputs: role `summary` (plain text), role `transcript` (SRT), and role `transcript-words` (JSON matching /.well-known/weftly-transcript-v2.schema.json, with segment-level and per-word timestamps). For other formats, pass `format=srt|txt|vtt|json|words|edit` to get_job_status to receive transcript content inline — `txt` and `vtt` are derived from SRT, `json` is v1 (segments only), `words` is v2 (segments + words), `edit` is a plain-text "[NNNN] words..." paragraph list for editing by hand. Flat price: audio $0.75, video $1.25 — see /.well-known/mpp.json for the authoritative table. Use for meetings, long-form interviews, lectures, and podcast episodes; the `words` output additionally supports creating clips, multicamera edits, or edit-video-from-transcript. Retrying any call with `job_id` alone returns current state (idempotent). Failed jobs auto-refund.

Transcribe audio or video to text, including per-word timestamps for precise editing. Three-call flow: (1) call with `filename` to receive {job_id, payment_challenge}; (2) pay via MPP, then call with `job_id` + `payment_credential` to receive {upload_url} (presigned PUT, 1h expiry); (3) PUT the bytes, then complete_upload(job_id), then poll get_job_status(job_id). On completion, get_job_status returns two outputs: role `transcript` (SRT) and role `transcript-words` (JSON matching /.well-known/weftly-transcript-v2.schema.json, with segment-level and per-word timestamps). For other formats, pass `format=srt|txt|vtt|json|words|edit` to get_job_status to receive content inline — `txt` and `vtt` are derived from SRT, `json` is v1 (segments only), `words` is v2 (segments + words), `edit` is a plain-text "[NNNN] words..." paragraph list for editing by hand. Flat price: audio $0.50, video $1.00 — see /.well-known/mpp.json for the authoritative table. Use for podcasts, interviews, meetings, lectures, and especially for creating clips, multicamera edits, or edit-video-from-transcript where word boundaries matter. Retrying any call with `job_id` alone returns current state (idempotent). Failed jobs auto-refund.