Clip Pack

The Clip Pack node cuts a podcast or interview recording into a pack of short clips in one step. It finds the best moments and renders each one as its own clip, framed for social and captioned. It can also cut a conversation filmed on several cameras.

Inside, the node runs the same steps that you can wire by hand: Transcribe with speaker labels and word timings, then Edit Plan in Clips mode. For each clip, it then runs a render, Speaker View with a framing or Apply EDL without one, and Add Captions when captions are on. Each step runs as its own job, so a long recording, a cancel or a failure behaves as it does in a hand-built workflow.

• The plan picks the moments. Edit Plan finds up to the number of moments set in Clips, each of about the Target length (seconds). It gives each moment a title and a hook. It may find fewer than you asked for. The pack holds the clips it found, and is charged for those renders only.
• Every clip is tried. A clip that cannot be rendered is skipped, and the pack keeps the others.
• Each clip is captioned on its own transcription, so the captions follow the clip's own words.

Clip Pack runs its planning step on Nodaro Cloud, and its render too when you choose a framing. On a self-hosted install, connect the install to Nodaro Cloud, under Integrations › nodaro.ai or with an API key. Without a connection, the node is not offered and a run is refused before anything starts.

When to use it

• Make a week of vertical, captioned clips for Instagram Reels, TikTok and YouTube Shorts from one episode.
• Make short audio snippets of an audio-only podcast to share as previews.
• Get a shortlist of moments, keep the best ones in the review, and render only those.

Use the individual nodes, such as Edit Plan, Apply EDL and Speaker View, when you need more control over a step.

Quick start

The template marketplace has two ready workflows. Clip Pack is the same chain built by hand, without a framing, ready to fill. Clip Pack in One Node is this node, with Review before render on, 8 clips of about 60 seconds framed 9:16 with stacked speakers, and karaoke captions.

Inputs

A recording that is longer than 3 hours, or a recording with no sound, is refused before anything is charged.

With more than one recording, the first one in the order of the Recordings list is Camera 1. The settings panel shows a Cameras section. See Several cameras.

Outputs

A clip whose captions could not be added is marked No captions on the node.

Settings

Framing

With Framing set to a shape, the node renders each clip with Speaker View on automatic boxes. Every clip is boxed the same way, over the episode's speakers. Those are the speakers that the transcript names who talk for at least 10 seconds, or for at least 5% of the episode's talk time. The framing needs two to six such speakers on one camera. Otherwise, the clips show the camera full frame.

• Layout offers only the layouts that Speaker View draws at the chosen shape.
• Switch sets what happens at each change of speaker.
• Emphasis marks who is talking: Scale, Border, Dim, any of them together, or none.

With None — the recording's shape, the node renders each clip with Apply EDL at the shape of the recording.

Following faces

With a framing chosen, Framing source sets where the speakers' boxes come from:

• Even split. The automatic boxes above: each speaker on screen gets an even share of the frame, with the same boxes on every clip.
• Follow faces. Speaker Frames tracks each face once, over the whole pack, and the Speaker View of every clip follows it. It is one run over every clip that the plan cut, never one per clip. Speaker Frames is charged per started block of 1,000 sampled frames. Faces are not matched to speakers yet: a camera that shows one person is followed, and a camera shared by several speakers is framed by the even split.
• Automatic (follow faces). Follows faces while Speaker Frames is available to you, and uses the even split otherwise, with no refusal and nothing extra charged. The estimate, and the listed price of an app or a template, follow the same rule.

On a self-hosted install, Speaker Frames needs a connection to Nodaro Cloud, as Speaker View does. Where Speaker Frames is not available to you, the settings panel offers no framing source, and Automatic (follow faces) is the even split. A node set to Follow faces then shows a note, and its run is refused with 403 node_not_available before anything is charged.

With Review before render on, the faces are tracked before the review, on the clips that the plan cut. Render final uses those tracks and does not run Speaker Frames again. On several cameras, Speaker Frames samples every camera of the switched clips, not the master audio.

Several cameras

Wire each camera's recording into Recordings. When you have the mixer or recorder track, wire it into Master audio. The run then lines the recordings up with Audio Sync, transcribes the sound, and works out which camera each speaker is on with Speaker Map. It plans the clips over every recording. For each clip, Camera Switch then puts each moment on whoever is talking, and the clip is rendered and captioned as it is for one recording.

• Recordings. 1 to 6 recordings of one conversation, in order: Camera 1, Camera 2, and so on. Drag them in the Recordings list of the settings panel to change the order. One of them may be sound only. Only one may be sound only, and none beside a master audio.
• Master audio. The sound of the pack, and the clock that the recordings are lined up to. Up to 6 recordings in all, the master audio included.
• Without a master audio. The sound is the wide camera's own sound, or else Camera 1's sound, so the sound never jumps between camera microphones. With two or more cameras, this keeps one place for the sound, so up to 5 cameras.
• One camera and a master audio. The camera is lined up to the master audio and cut on that one camera. A master audio wired from the same node as one of the cameras is that camera's own sound, and the node runs as without one.
• Wide camera. The camera that shows everyone. A speaker with no camera of their own is shown on it, and it is never a speaker's camera. None by default.
• Recordings already in sync. Turn it on when every recording already shares one clock. The run then skips the line-up, so there is no Audio Sync step.
• Who is on which camera. By default, Speaker Map decides from the cameras' own microphones: the camera that hears a speaker loudest is theirs. A speaker that no camera clearly leads for has no clear camera, and is shown on the wide camera, or else on any camera that filmed them. The node then says so under its result. Set a speaker to a camera, or to the wide shot, to override the automatic choice. Speaker 1 is the first voice heard, Speaker 2 the next, and so on.
• Every clip is tried. Camera Switch runs once per clip. A clip whose camera switch fails is skipped and not charged, like a clip that fails to render. The pack keeps the other clips.
• If a multicam step fails. Audio Sync and Speaker Map run before every clip. When one of them fails, no clip is made and the run fails, and its per-run fee is refunded. Continue runs the step that failed and the steps after it.
• Review before render. On several cameras, Render final of a reviewed pack runs Camera Switch again on each clip you kept, then its render and its captions.

The run's result names who was put on which camera, with the notes of the line-up and of the map. Where Speaker Map is not available to you, a run in the editor or in an app reads only the first recording, and the settings panel says so.

These are refused before anything is charged:

• Two sound-only recordings, or one sound-only recording with a master audio.
• More recordings than the limits above.
• A wide camera that is not one of the recordings with picture.
• Settings under Cameras that are not valid.
• A master audio that cannot be read as audio.

Cameras settings on a run of one recording with no master audio are ignored.

Review before render

With Review before render on, a run stops at a preview of each clip. Each render is a Preview, at the lower preview rate, private and at 720p. The captions are not added yet.

1. Open the review. It shows every clip as a card. Keep or drop each clip, and edit its hook.
2. Click Render final on the node. The node renders only the clips you kept, at full quality, and captions each one, once.

• Keep at least one clip. When every clip is dropped, the node shows "Keep at least one clip to render." and Render final is disabled.
• Render final uses the settings of the previews. To change a setting, such as the framing, run the node again.
• You can render final again. After a Render final has finished, edit the review again and click Render final. The renders and captions of the kept clips run and are charged again, with no fee. Rendering the same edit a second time is refused, and only one Render final runs at a time.
• In a published app, the node shows its previews on the app's page with a Render final button, which renders every clip at full quality. The app's page has no review, so it drops no clip. Once a final is made, another Render final of the same run is refused. See Render final in an app.
• The nodes after it run on the previews. On Nodaro Cloud, the node stops at its previews, but the nodes connected after it run on the previews and are charged. They run again on the final clips at Render final.
• Someone has to review. A run with nobody to review it cannot pass the node. That includes a run started from the API, the SDK, the MCP run_workflow tool or a trigger, and a run inside a component. The nodes before Clip Pack run and are charged. The node then fails with preview_review_required, with nothing charged for it. Turn Review before render off for those runs.

The MCP tools make_clip_pack and continue_use_case are reviewed by the AI assistant that calls them, so they can use Review before render.

If a run fails

• A clip that could not be rendered, or, on several cameras, a clip whose camera switch failed. The run completes with the clips that rendered and keeps its fee. The node says how many could not be rendered, for example "1 clip could not be rendered." A clip that failed is not charged. The run completed, so it has no Continue. There are two ways to get the missing clip:
  • The same clip, with no fee. Render it yourself from the Clip set (EDLs) output. Wire it into an Apply EDL node, or a Speaker View node for the framed look, with the connection's selection set to that clip. Add an Add Captions node if you want captions. Each step is charged at its own rate.
  • A new run. Run the node again. Every step is charged again, including the per-run fee, and the new plan may pick other moments.
• Captions that could not be added to a clip. The clip is delivered without captions and marked No captions. Captions that failed are not charged.
• No clip rendered. The run fails, and its per-run fee is refunded. The message gives the count and the reason, for example "None of the 8 clips could be rendered: …".
• The plan found no clips. The run fails with "The plan found no clips in this recording." Nothing is rendered, and the per-run fee is refunded.
• A run that stopped before its captions ran. The run fails, and its per-run fee is refunded. Press Continue: it captions the clips already rendered, without rendering them again.
• Any other failed run. The steps that finished stay finished. Their results, such as the transcript, the plan and the clips already rendered, are kept and are not charged again. Press Continue on the node: only the step that failed and the steps after it run again.

Not once per list item

A Clip Pack takes the recording of one conversation, and its result is already a list, one entry per clip. A Clip Pack fed by a List is refused before any of its runs starts, with nothing charged. The message reads: "Clip Pack takes the recordings of one conversation. Use a separate workflow run for each episode."

Credits

Clip Pack charges each step at its own rate, the same as the node it stands for, plus a small per-run fee. The node shows the full estimate, fee included, before you run.

• The per-run fee is charged once per run. A Render final after a review is not charged the fee again. A failed run's fee is refunded, and Continue holds it again. A run that keeps some of its clips is charged the fee.
• You are charged for each clip's own length, not for the estimate. Each clip's render is charged for that clip's minutes, rounded up on their own, at that clip's rate. A clip that was not rendered is not charged for its render, and a clip without captions is not charged for them.

The estimate before you run

The estimate prices each clip at twice the Target length (seconds), so it leans high. At the default 60 seconds, each clip's render is quoted at 2 minutes. With a framing, Speaker View is quoted at its higher rate, previews and Render final included. The one exception is a Single layout with no zoom switch, which is quoted at the lower rate. The estimate counts the clips you asked for, although the plan may find fewer. After the previews, Render final is quoted on the clips that you kept, each at its own length. A recording with no known length is estimated at Edit Plan's 180-minute ceiling. How Edit Plan counts the minutes is explained on Edit Plan.

Nodes after Clip Pack

A node wired to the Clips output on an Each connection runs once per clip, and is charged once per clip. For example, a Social Media Format node after an 8-clip pack is charged eight times. The estimate counts the clips you asked for. The run charges the clips that were delivered.

Tips

• Steer the moments. Use Instructions, for example the funniest moments or only the part about pricing.
• Drop the weak clips first. Turn on Review before render to drop weak clips and sharpen their hooks before anything renders at full quality.
• Post each clip once. Wire Clips into a publishing node on an Each connection.

Troubleshooting

From the API

POST /v1/clip-pack starts a run and answers 202 with jobId and executionId. The body takes mediaUrls (one recording URL, or up to six of one conversation in order), and optionally masterAudioUrl, cameras, count, targetDurationSec, framing, framingSource, captions, captionStyle, reviewBeforeRender, planTier, instructions and styleGuide. Poll the job with GET /v1/jobs/:id/status.

• framing is { aspect, layout, switchType, emphasisStyle }. Leave it out for the default (9:16, stacked, pan, scale). A framing without a layout is stacked, moved to a layout that the shape draws (side by side at 16:9). null renders the recording's shape. An emphasisStyle that the node does not draw answers 400 validation_error.
• framingSource is "auto", "static" or "tracking". Left out, it is "static", the even split. "tracking" is refused with 403 node_not_available where Speaker Frames is not available to you.
• cameras is { wide?, synced?, speakerCameras? }. wide is the recording number of the wide camera, or null. synced: true skips the line-up. speakerCameras maps a speaker number to a recording number, where 0 is the wide shot. See Several cameras.
• Several recordings, or a masterAudioUrl, need Speaker Map. Where Speaker Map is not available to you, they are refused with 403 node_not_available before anything is charged.
• The job's output_data carries the clips (clips, a list of { url, title, hook, clipKey, thumbnailUrl, captionsFailed }, in the plan's order) and the first one as videoUrl (audioUrl for audio clips). failedClips counts the clips that could not be rendered, a clip whose camera switch failed included, and failedCaptions the clips without captions, each only when there are some. It also carries the clip set (json), the transcript (transcript) and quality (proxy for previews, final otherwise). A run on several recordings or a master audio adds cameras: each speaker, the recording number of the camera it got (null for the wide shot), and the notes of the line-up and of the map. A failed run keeps what completed and names the failed step in failedStep.
• An Idempotency-Key header makes a retry of the same request answer with the same jobId and "deduped": true, and start nothing.
• POST /v1/clip-pack/estimate takes the same body and answers { data: { totalCredits, stepsCredits, feeCredits, durationSec } }, without starting anything.
• POST /v1/clip-pack/continue with { jobId, mode } continues a run. "mode": "from-failed" runs what a failed run did not complete. "mode": "render-final" renders a review run's kept clips, with an optional editedEdl that keeps or drops clips and edits their hooks. A review that drops every clip answers 409 use_case_nothing_kept. A Render final of a run that was not a preview answers render_final_not_preview, and one of an edit already rendered answers render_final_already_rendered.

The SDK (client.edit.clipPack, client.edit.estimateClipPack, client.edit.continueUseCase), the CLI (nodaro edit clip-pack, with --estimate for a quote, and nodaro edit continue-use-case) and the MCP tools make_clip_pack, estimate_use_case and continue_use_case call the same routes. See Run a single node, Editing in the SDK and the MCP analysis and EDL tools.