How to license video to AI labs: the full process

First, describe what you actually have. “Video” could mean unused camera originals, edited films, first-person recordings of work, road footage or production projects with linked assets. Each presents a different licensing and preparation question.

First-person, or egocentric, video is filmed from the wearer’s viewpoint. A camera recording does not automatically include the force, motion or control signals a robotics buyer might need. If the buyer needs those signals and you never recorded them, discuss a separate capture project. Renaming your existing footage cannot supply them.

For an existing collection, the process looks like this:

In words: establish fit, agree what may happen, test an authorized sample, identify the licensed delivery, then track the commercial events separately. A single agreement can cover several of these functions. The important point is that each operation is permitted before it happens. An existing grant does not need a second approval merely because it appears at another stage of this diagram. Payment may also be upfront or staged if the agreement says so.

Assign a person to each handoff. Otherwise the rights review, technical review and commercial decision can become three parallel conversations that never meet.

The accompanying blank and worked records cover the full process. Keep the real evidence in controlled storage and use references in the worksheets.

Your rights lead and media custodian may be different people. So may the buyer’s technical evaluator and authorized signatory. A broker’s involvement does not itself establish authority to license the footage, operate on it or receive the owner’s money.

Here is a useful first description, using the invented collection:

Invented initial scoping note: Workshop A contains workshop camera originals and edited exports. We are separating duplicate edits and reviewing people, background audio and third-party inserts. Please confirm the intended AI use, the footage unit you would evaluate, and the minimum non-sensitive information needed to assess fit. We have not authorized access to the footage at this stage.

You can write that note before preparing clips. Keep private storage links, credentials and sensitive location details out of it.

A broad demand category is a starting point for a conversation. It does not tell you that a buyer has approved your material or committed a budget. When checked on November 15, 2026, Rightmark’s Wanted page showed video categories but no individual open request. Use those categories to frame questions, rather than treating them as purchase orders.

A public listing is more specific, but still has limits. Rightmark’s private-sale catalog includes a road-video lot with an owner-estimated 150 to 170 hours, offered non-exclusively and priced on review. The proposed scope excludes audio, trims location information and requires agreed treatment of incidental people and plates before delivery. The listing also flags possible minors. This is an owner’s proposed collection, not an audited volume, an accepted delivery or evidence of payment. Its existence does not override Rightmark’s restrictions on data about children.

The four priced examples in step 4 show why the footage type matters: aerial views, first-person tasks, synchronized driving recordings and finished animation each have a different asking scope. They are offered collections, not a ranking of lab demand. Keep a proposal to record new work separate from an existing archive: future capture capacity supplies neither existing accepted hours nor a price for them.

I would use four different descriptions in a deal file: a demand category, an owner-described collection, a scoped buyer inquiry and a signed license. Delivery, acceptance and payment then need their own records. That vocabulary prevents “the lab is interested” from becoming “the lab bought it” in an internal forecast.

Define the hour before negotiating the hourly rate

Stored file duration, unique source duration, candidate footage and accepted footage answer different questions. Agree which one determines the invoice.

Two exports of the same recording can double your file hours without adding a new event. Overlapping clips can count the same seconds twice. Two cameras may provide valuable distinct views, but the buyer needs to say whether it pays for each camera-hour or for the underlying event once.

An established dataset illustrates why identification matters. Ego4D’s video documentation distinguishes source components, canonical videos and clips with their own timelines. Use that distinction to ask what your own identifiers mean. It is not a requirement to copy Ego4D’s formats or frame rate.

For the invented Workshop A collection, I would propose this unit card:

That end-time convention avoids an ambiguity at adjoining clip boundaries. The rest matters more commercially: a buyer who needs a complete action may reject a high-resolution clip that stops before the outcome.

Now reconcile the inventory. These invented deductions are disjoint: each duration is deducted once, even if it has several problems.

The last 60 hours still need the remaining rights, preparation and buyer checks. The subtraction has not cleared them for licensing. Use the quantity worksheet to preserve each deduction and its reason.

Keep clip counts in perspective, too. Invented arithmetic: 1,200 non-overlapping clips of 30 seconds total 10 hours. Sampling them at one frame per second produces 36,000 frames. It creates a different representation of the same footage, not another 10 hours of observation or 36,000 independent events.

The next fit note can therefore be precise: Workshop A has 60 candidate hours under the proposed counting rule, no audio or sensor delivery, and unresolved permissions to close. Ask whether the buyer needs the recorded views and existing labels. If it needs a signal or event you do not have, narrow the offer, propose a separately priced capture job or stop before funding preparation.

Start with the entity that would grant the license and the person authorized to sign. Then trace how that entity obtained the relevant rights: employees, contractors, co-producers, clients, acquisitions and previous assignments can all change the answer.

As a limited US reference, the Copyright Office explains ownership through authorship, qualifying work made for hire and transfers. Commissioning a video does not by itself establish that every component falls into the same ownership category. Use the actual agreements and the applicable jurisdiction to resolve your collection.

Review the components separately. You may own a camera recording while an inserted clip, soundtrack or stock asset has restricted onward use. A final render and the editable project may have different dependencies. Permission to publish on a channel or deliver a client film also needs to be checked against the proposed AI use.

Rightmark’s Trust policy requires consent naming AI training for people’s voices or faces and says it does not accept data about children. Those are relevant limits for the Rightmark route. They do not resolve every legal question in every jurisdiction.

Keep ownership, privacy and usefulness in the same review

A person may be identifiable through a face, voice, name badge, screen, vehicle plate or surrounding context. Logos, premises and property can also raise contractual or other questions depending on the footage; a universal “release required for every object” rule would not help you resolve them.

For processing within the GDPR’s scope, the EDPB’s video-device guidance, adopted January 29, 2020 and corrected February 26, distinguishes collection from disclosure. Disclosure requires its own legal basis; a new purpose needs assessment, and the recipient must analyze its own processing. The guidance also explains why an employee’s consent may not be freely given where there is a power imbalance. An old workplace recording notice therefore deserves specific review before a new AI licensing use.

Here is what that review might find in invented Workshop A:

These are issues to investigate, not a declaration that the remaining collection is cleared. They also are not six additional hour deductions from the quantity bridge. Keep multiple reasons against an interval without subtracting its duration repeatedly.

Blurring can change the footage’s usefulness. If the buyer needs to follow a hand entering a code on a screen, masking the screen may remove the event it wanted. If a mask leaves a name badge or recognizable voice, the privacy question may remain. Decide what must be removed and what must still be visible, then test both in the prepared copy.

In our invented monitor example, the clip stays on hold because neither side has yet established a permitted treatment that preserves the required action. Calling the output “anonymized” would not answer that problem. Record the source, treatment, output version, unresolved exceptions and reviewer. Keep any link back to restricted originals under the appropriate controls and retention rules.

Your ownership pack should point to the evidence, authorized signatory, existing grants, exclusions and proposed use. Rightmark’s provenance log provides a way to reference underlying records without placing private documents in a public worksheet. A well-organized pack makes review possible; it does not itself confer rights or legal clearance.

A useful estimate begins with a comparable unit and a defined grant. Before multiplying anything, ask whether a figure describes unused creator footage, task-oriented first-person video, undifferentiated bulk footage or a newly commissioned recording service. These are different purchases.

Rightmark’s Q4 2026 Price Index, published October 26 and rechecked November 15, contains the following reference points. Its confidence labels describe the evidence, not the chance your footage will sell.

Sources: The Decoder’s report, Stellaris’s analysis, Flikforge’s estimate.

Do not average these rows or treat one as a price floor for another. They have different footage, units, dates and evidence behind them. None establishes an accepted tariff for your collection or your net payout.

Indicative estimate, not an offer. Final offers depend on the systems you connect, data size and quality, access terms and due diligence.

For scale only, 100 invented minutes multiplied by the historical creator-footage range produces $100 to $400. That arithmetic says nothing about whether a buyer wants those minutes, the rights being granted, preparation costs or the owner’s proceeds.

Read a public ask before comparing a buyer’s proposal

I would move through the price discussion in this order: choose a dated reference for the right footage and unit, obtain a description-based indicative appraisal, then review a permitted sample and the proposed scope before comparing a buyer’s proposal. The valuation guide distinguishes the initial range from a route-specific quote after review. Put the agreed price, licensed units, fees, acceptance and payment conditions in the signed terms. An NDA addresses confidentiality; the sample operations still need permission.

Here are four published asking scopes, checked on November 15, 2026. Each listing is open for interest and requires owner confirmation before samples and terms. The quantities and status labels are the listings’ descriptions, not our independent measurements or buyer acceptance. The confidence label describes the evidence behind the indication; it is not a probability of sale or a rights audit.

None of these asks records a buyer-agreed price or completed sale. Do not divide the package amounts by hours or minutes that have not been confirmed as the licensed, accepted scope. The first-person lot here is AV-EGO, not the separate AV-EGO-LABELED lot; the aerial lot is AV-DRONE, not AV-DRONE-UHD.

Indicative estimate, not an offer. Final offers depend on the systems you connect, data size and quality, access terms and due diligence.

The commercial discussion can change the quantity, permitted uses, preparation, exclusivity and payment conditions together. The invented agreement below shows how to compare a defined bargain. It does not assign that bargain to any of these real collections.

Price the accepted scope, then calculate both sides’ costs

Return to invented Workshop A. For this calculation, assume that the necessary permissions and preparation have subsequently been resolved, the parties agree $30 per accepted hour, and their contract expressly permits independently useful partial acceptance.

All 60 hours accepted would produce a $1,800 license payment. If only 54 hours meet the agreed criteria, the license payment would be $1,620. If the buyer instead contracted for one indivisible package, neither calculation establishes a right to invoice a subset. We will return to that distinction at acceptance.

Rightmark’s current published fee model says owners pay nothing upfront. On standard licenses, Rightmark takes nothing from the owner’s payout and the buyer pays its fee on top. On premium and bespoke deals, the owner keeps 75% to 85%, agreed in writing before a sale. The page also says buyers cover extraction, scrubbing and transfer. It does not publish a universal buyer-fee percentage.

Ask for a written fee schedule that identifies the payer, calculation base, rate, trigger, included work and adjustments. An introduction, preparation service, negotiation mandate and payment-collection role are different jobs. A scout’s share of Rightmark’s income should not quietly become an additional owner deduction.

The following invented standard-style arrangement uses a 10% buyer fee solely to make the arithmetic visible. It is not a quoted Rightmark rate. The buyer pays the owner directly, and the assumed broker fee is due with that payment. Preparation costs $600, transfer costs $60 and technical review including staff time costs $360; the buyer owes those amounts in either outcome. The owner pays an adviser $100 and spends four hours of administration valued at $40 an hour.

The owner contribution is not accounting profit. This model assumes all stated cash amounts settle and includes no other storage, tax, withholding, currency, refund or financing costs. Replace those assumptions with actual amounts or explicit unknowns when comparing your offers. Do not turn an unknown cost into zero. The blank offer worksheet and filled cost model let you separate those inputs.

The buyer’s cost per accepted hour rises in the second column because the fixed work is spread over less accepted footage. That can explain a buyer’s concern about an uncertain archive even when its advertised license rate looks reasonable.

A buyer-paid fee can also affect negotiation through the buyer’s budget. In a separate invented fixed-budget example, the buyer has $1,800 for the license plus a 10% fee calculated on that license. The available license amount is $1,800 ÷ 1.10 = $1,636.36, with $163.64 for the fee. That is a budget negotiation before agreement, not permission to deduct a fee from an already agreed $1,800 owner payment.

For a different invented revenue-share arrangement, an expressly defined $1,800 collected base split 80/20 gives the owner $1,440 and the intermediary $360. Do not add the earlier 10% fee to this alternative unless the actual terms require both. Define the base, adjustments and collection trigger. The example’s 80% is an assumption, not a standard promise.

Find out why an offer is low before spending to improve it

Ask the buyer which limitation is reducing the offer: task fit, eligible duration, rights, duplicates, preparation, permitted uses or available budget. Then ask what observable change would affect its decision and who would pay for that change.

A higher rate can be a worse deal. In an invented upgrade offer, an owner can receive $1,200 for a raw permitted collection or $1,500 after adding labels. If the owner must spend $500 on those labels, the upgrade leaves $200 less cash before other differences. Negotiate preparation funding or a narrower specification before commissioning the work.

I would improve value by resolving a specific purchasing obstacle: establish the useful subset, reconcile the counting rule, locate the missing rights record, preserve a task sequence, or supply labels the buyer has actually requested. Adding expensive annotations without that agreement can create an impressive package for a different buyer.

Compare exclusivity with the opportunities it removes

Define the footage, AI use, territory, recipients, start event and term covered by exclusivity. Record retained uses, previous grants and whether later recordings are included. An exclusive sales channel is also a different restriction from an exclusive right to use the footage.

In an invented comparison, a non-exclusive license pays $1,800. A 12-month exclusive license for a narrowly defined AI-training use pays $2,400. The premium is $600. Suppose one potential future license within the restricted scope would contribute $900 net. At assumed probabilities of 0%, 50% and 100%, the expected contribution forgone would be $0, $450 and $900 respectively.

That is sensitivity analysis, not a prediction that the future sale exists. Add other plausible uses, the reliability and timing of payment, and the burden of policing the restriction. A broad grant can close opportunities beyond the one you happened to model. Rightmark’s valuation guidance makes the same essential point: exclusivity alone does not make payment larger, faster or more certain.

If you are comparing an offer now, use the Rightmark Price Index to check its unit and evidence date, then put the owner’s net proceeds and buyer’s full scoped cost beside the rate.

Keep the original under the controls and retention rules that apply to it. Prepare an authorized working copy with stable source identifiers. Record durations, streams, resolution, timing information and known capture dates, then connect each output to its source. Unknown values should stay unknown until measured.

A local media report can help with this inventory. If you already have an approved media tool, use its report. For a technical operator with FFmpeg installed, this example uses ffprobe’s documented field selection to return selected media properties as JSON:

ffprobe -v error \
  -show_entries format=duration,size:stream=index,codec_type,codec_name,width,height,avg_frame_rate,r_frame_rate,time_base,start_time,duration \
  -of json "/approved-local-copy/clip.mp4"

Run it only against a copy you are authorized to inspect. A missing field is not zero. This inventory command does not prove that every frame decodes, that the footage is unique, that a task is visible or that you have rights to license it. It also is not a complete privacy scan: filenames, tags, other metadata and sidecar files still need appropriate review.

Agree the required codec, resolution, orientation, frame-rate handling, audio treatment, labels and synchronization with the buyer. A technically valid conversion can still break a useful relationship.

Check what the cleaning changes

Consider this invented timing failure. A source recording contains a task at seconds 110 to 118. The preparation job extracts source interval [100,130), producing a 30-second clip. The first label file still points to seconds 110 to 118 in that new clip. It cannot describe the intended event there.

Subtracting 100 works for this simple trim. More complicated edits need an actual transformation map, including gaps or changed timing. An arithmetic correction alone does not prove the event remains visible.

Practical check: validate labels against the delivered version, not a familiar original. If the scope excludes audio, inspect the output streams as well. A preview player with its volume muted does not demonstrate that the file contains no audio track.

For Workshop A, the invented work order WA-PREP-01 can be specific enough to commission:

This work order does not accept the whole collection. It defines a preparation deliverable. If a test reveals that the original does not contain the necessary action, the next decision may be a scope change rather than another edit.

Specify the custody arrangement behind a “clean room”

Write down the purpose, authorizing grant, custodian, permitted operators, tools, environment, logging and outputs. Say which copies may exist, how long they may remain and what ends access. If an operator proposes a buyer-supplied executable or an external API, establish what that tool reads, sends and retains before permitting it to run.

The phrase “clean room” does not tell you whether an environment has network access or whether an operator can export frames. Those controls need to be specified and implemented. For video, also decide whether a reviewer can retain thumbnails, screen recordings, transcripts or model-derived representations.

Rightmark’s current Clean Room page describes operational-archive early access and distinguishes it from Repo Check’s available code mode. It does not establish a generally available automated video-cleaning pipeline. Scope the actual video environment, operator and checks for the proposed deal rather than importing a code or archive capability into your offer.

Start with the question the buyer cannot answer from the description. A schema check may need a small technical specimen. A task-visibility check needs relevant footage. A proposed model experiment may require different permissions and a different environment. Choose the least revealing evidence that can answer the actual question; even a contact sheet discloses content.

Explain how you selected the sample. A deliberately chosen set can demonstrate a format or expose a specific failure. It cannot establish a defect rate for material excluded from that selection.

For invented sample WA-S1, assume the relevant permissions have now been obtained for 12 distinct 30-second clips. Four show assembly, four show a tool change and four contain partly occluded actions. Together they contain six minutes of video. The package includes a source map, matching labels, a data card and known gaps. It contains no audio, sensors or held material.

The question is whether the buyer can follow the actions and match them to the labels. The owner selected the clips; they are not a random draw. That selection can help the buyer decide whether to continue. It cannot prove that all 60 candidate hours have the same characteristics.

Match the evaluation to the actual permission

Record the sample version, recipients, permitted operations, environment, retention period and allowed outputs. Confidentiality and use are separate questions. An NDA’s confidentiality protection does not by itself authorize AI training, onward sharing or every test a recipient would like to perform. The actual agreement may already cover several functions; read its grant rather than relying on its title.

In the invented WA-S1 permission, the named buyer may manually inspect the clips in the agreed environment and retain a report of identifiers and results under the stated terms. It may not update model parameters, create reusable embeddings, use an external service or pass the sample onward. The retained report may not contain frame excerpts.

Suppose the evaluator proposes uploading clips to an external vision API. That proposed operation stays on hold while the parties resolve permission and processing terms. The already permitted manual review can continue. A service’s “no training” setting does not itself authorize transferring someone else’s footage to it.

This is why I would ask for the evaluation plan before sending a sample. An otherwise useful buyer can have a perfectly reasonable test that your current permission does not cover. Finding that out early gives both sides a concrete choice: authorize a suitable arrangement, use a different test or decline the proposed evaluation.

Turn the result into a decision, not a vague approval

In our invented sample result, all 12 files parse and their identifiers and timing references resolve. One clip fails the visibility criterion because an object blocks the action. Record that failure against the clip, version and criterion. Do not extrapolate a one-in-12 failure rate to the archive.

The next request might be a permitted alternative view or confirmation that no suitable view exists. If the owner supplies WA-S2, test the changed item and affected outputs in that version. A repaired sample needs its own result; it does not prove that the full collection has been regenerated or corrected.

Use a sample log with these fields: package version, selection method, excluded groups, permission reference, question, criterion, result, evidence, decision-maker and next permitted action. Include the deadline and what happens to copies and outputs when review ends. The sample worksheet keeps these fields together. That record is more useful than “sample approved,” which could mean format compatibility, task fit or agreement to proceed commercially.

Put the first-sample estimate on the correct clock

Rightmark’s buyer page currently gives video and imagery an indicative 2 to 6 weeks from the end of scoping to first samples. It labels this an October 2026 planning estimate at Low confidence, with timing confirmed at scoping. It is not a promise from first inquiry to payment.

The complete Workshop A schedule in step 8 uses an invented 14-calendar-day preparation interval before the first sample. Choosing a duration inside that published range does not make it typical or turn the other milestones into observed practice. The sample’s visibility failure remains a failure; the schedule assumes no successful replacement.

For your collection, record when scoping is complete, when permission is effective, what preparation remains and who can start the buyer review. Agree any pause conditions and a response deadline. The commercial negotiation, delivery review, invoice and payment clocks still need their own terms.

By signing, both sides should be able to identify the collection and describe the deal without relying on an informal call. Identify the licensor, authorized signatory and licensee. If a partner has delegated authority to license or collect payment, record that authority rather than inferring it from an introduction.

Name the version, units, streams, labels and supporting files. Then define the operations the buyer may perform: inspection, evaluation, training, frame extraction, annotation, embedding creation, deployment or onward licensing may require different treatment. Name the relevant recipients, affiliates and processors. A broad phrase such as “AI use” leaves too much for an engineer to interpret alone.

Align the license with the sample permission and preparation work order. If the final grant already permits a specified transformation, there is no need to invent another approval step for it. Equally, a permission to test six minutes should not silently become permission to train on the whole collection.

Here is an invented pre-signing scope record for Workshop A. Its open cells are decisions to resolve, not standard terms:

This record is not ready to sign. A useful term sheet makes those gaps visible while the parties can still change the bargain.

Decide whether a usable part can be accepted independently

A collection may be useful clip by clip, or only as a connected whole. If the buyer needs a continuous sequence with synchronized labels, several playable files may not fulfill the task. A common rights defect can also affect every clip, even if each passes a technical test.

For independent partial acceptance, connect three things in writing: the usable subset, the amount allocated to it and the condition that permits invoicing that amount. For one dependent package, explain what the full package must do. Neither side should discover after a failed delivery that it assumed the other structure.

For the same invented 54-of-60-hour result, the agreed structure changes the invoice decision:

Write the acceptance procedure with the same care as the specification. Which event starts review? Who can accept or reject? What evidence accompanies a rejection? What happens if there is no response? Do not invent automatic acceptance merely because a review date has passed.

Correction, replacement, refund, termination and additional preparation costs also need agreed treatment. A defective export, missing original information and a newly requested label are different problems. Specify who must respond, who pays, the relevant deadline and whether the change affects a review clock. Rightmark does not provide one universal remedy or payment trigger for every video deal through this guide.

The acceptance and payment guide develops the same distinction between an identified delivery, an accepted scope and an amount that may be invoiced. Apply it to your task dependencies before agreeing a staged payment plan.

Say what survives when access ends

Review more than the original clips. Extracted frames, thumbnails, labels, embeddings, review reports, caches, model weights, outputs, backups and audit records can each have a different permitted life.

For each relevant class, identify the recipient, allowed use, retention rule, end event and required action. A short access term does not by itself establish that a trained model must be deleted. Nor does disabling a download link recall copies already taken. If deletion, restricted retention or a confirmation is required, describe the obligation and its limits; do not assume automatic model unlearning.

For invented WA-S1, the retained report is limited to clip identifiers and evaluation results under its stated terms; frame excerpts and reusable embeddings are outside that sample permission. The later license’s model-retention row remains unresolved in the pre-signing record above. Keeping those two rows separate prevents a sample exception from deciding the final model rights by accident.

Finally, check the warranties, liability allocation, publicity, audit rights, termination and governing-law provisions appropriate to the deal. Examine broker or channel restrictions separately: protected introductions, non-circumvention and fee tails can affect later dealings even where the data license is non-exclusive. Record the actual trigger and end event. A tail measured from the end of an appointment is not measured from the day of an introduction.

Before signing, I would ask the delivery operator to locate six answers in the documents: what moves, who may use it and how, what constitutes acceptance, what permits an invoice, what starts payment or onward remittance, and what must happen when access or the license ends. If those answers cannot be found, the operator is being asked to invent the process.

Once the applicable permissions and terms are effective, freeze the authorized delivery version. Package its manifest, data card, source map, label schema, exclusions and check report. Reference controlled rights evidence without putting unnecessary personal or confidential documents into the media bundle.

A practical manifest needs to identify the collection and delivery version; each file’s ID, relative path, type, byte count and checksum; duration, codec, resolution and timing properties; source intervals and transformations; audio and sensor status; label references; checks and relevant evidence references. Use the blank delivery manifest as a starting point and adapt it to the agreed delivery.

Here is a compact invented specimen:

The checksum proves file identity relative to the recorded checksum, not rights, safety or task quality. A correctly transferred file can still be the wrong licensed version.

Separate receipt from the decision to accept

Use the agreed transfer method, scoped access and retention rules. Record when the complete package is available and when the recipient acknowledges it. At receipt, reconcile filenames, sizes and digests, inspect unexpected files, and run the agreed decode, label and task checks against the delivered version.

A defect record should identify the criterion, file or interval, version, evidence, responsible role and proposed next action. Separate a transfer error from a preparation failure, a new requirement and a rights issue. They may have different remedies and funding arrangements.

If the package changes, issue a new version and rerun the affected checks. Record what changed and which earlier results still apply. Do not silently overwrite a delivery and preserve an acceptance result for a different set of files. Whether a correction restarts review depends on the terms; the filename does not decide it.

For this invented continuation, assume the parties have resolved every open item in the earlier term sheet and signed the necessary rights, privacy, use and commercial terms. They expressly allow independently useful partial acceptance and agree that accepted clips satisfy the required coverage without a common rights defect.

The buyer accepts 54 of 60 hours. Six fail the visibility criterion and remain outside the accepted use and payment scope while the parties consider the agreed remedy. The license price eligible to be invoiced is 54 × $30 = $1,620. At the moment acceptance is recorded, the owner has received $0 from this invoice. The acceptance decision establishes an entitlement under these invented terms; it is not a bank receipt.

If the contract had required one complete dependent package, this result would need a different decision. Do not apply the 54-hour calculation to that arrangement without an agreed change.

Put each deadline beside its trigger

There is no single useful “licensing takes X weeks” answer without defining the beginning and end. Use a schedule like this:

Here is a complete invented inquiry-to-cash plan for Workshop A. The company, buyer, dates, durations, agreement and payment are invented. All times are UTC; scheduled business days are Monday through Friday, with no holidays, including dates that would be holidays in real jurisdictions. Real staffing, calendars and prerequisites change a plan.

The rights review and negotiation windows overlap other work; adding every row’s duration would double-count time. Only specifically permitted sample work proceeds while other rights questions remain open. A date on the calendar cannot release held footage. The proposed external-API test remains outside the sample grant even while permitted manual review continues.

The plan assumes no additional delaying hold or newly paid scope beyond the stated work. Missing permission, inputs, useful footage, signatories or a required check pauses the dependent work. Record the actual pause and restart rule, revise affected dates and recalculate the total. The blank process schedule and filled fictional schedule keep prerequisites and parallel work visible.

Continue the same invented plan through delivery review, invoicing and payment. The agreement excludes the starting day when counting business days. Its review period is five business days from complete delivery; payment is due 30 calendar days after receipt of a conforming invoice.

The complete direct-payment scenario runs from November 30, 2026 at 09:00 to February 19, 2027 at 09:00: 81 elapsed calendar days, or 11 weeks and 4 days. The earlier stages are explicit planning assumptions, not a typical video timeline, forecast, observed deal or upper limit. The parallel work is already inside that total. Changing the assumptions can move both the milestones and the total.

If an authorized intermediary collects the license payment instead, add that arrangement’s remittance clock. In a separate invented collection alternative, cleared funds arrive with the collector on February 19 at 09:00 and remittance is due seven business days later, March 2 at 17:00, under the same weekday convention. That seven-day term is invented. It is not a Rightmark standard or an additional delay in the direct-payment example.

Reconcile the final amounts and close the permitted access

For the invented direct-payment outcome, the owner’s $1,620 cleared receipt less the $100 adviser cost leaves $1,520 cash. Subtracting the modeled $160 of administration time gives $1,360 contribution within the stated assumptions. The buyer’s $1,620 license payment, $162 fee and $1,020 preparation, transfer and review costs total $2,802. The six unaccepted hours are not automatically an additional debt.

Keep the acceptance record, invoice, payment evidence and any authorized remittance statement aligned to the same scope. Record the remaining six hours’ status and any separately agreed correction. Do not replace them with new footage merely because that would make the volume add up.

Close access and handle retained artifacts according to the actual terms and applicable recordkeeping duties. Keep enough lawful history to understand the grant, prior licenses and delivery decisions. An expired sample link, a deleted working copy and the end of a trained model’s permitted use are different events; record the ones the agreement actually requires.

I would start with one collection, not the whole archive. Its value becomes easier to discuss when you can connect the buyer’s task to identifiable, permitted footage and then connect the offer to the money you keep.

Use a short next-decision record: collection and version; current stage; proposed use; unit and quantity status; evidence already available; the open question; the person who can decide it; the next permitted action; and the event that starts its deadline.

For invented Workshop A at the initial stage, that record would say: 60 candidate hours under the proposed counting rule, rights still to resolve, no files released, and a request for the buyer to confirm task fit and the minimum sample question. It would not say “60 hours sold.” Later versions can carry the actual permissions, checks and commercial decisions as they occur.

• Worksheet instructions (Markdown)
• Acceptance events settlement blank (CSV)
• Agreement scope blank (CSV)
• Data card blank (CSV)
• Delivery manifest blank (CSV)
• Derivatives and next decision blank (CSV)
• Full process schedule blank (CSV)
• Invented acceptance settlement (CSV)
• Invented agreement scope (CSV)
• Invented data card (CSV)

• Invented delivery manifest full (CSV)
• Invented derivatives and next decision (CSV)
• Invented economics (CSV)
• Invented events (CSV)
• Invented full process schedule (CSV)
• Invented manifest specimen (CSV)
• Invented offer alternatives (CSV)
• Invented preparation custody (CSV)
• Invented quantity bridge (CSV)
• Invented rights register (CSV)

• Invented sample result (CSV)
• Invented scoping and unit card (CSV)
• Invented timing map (CSV)
• Inventory rights blank (CSV)
• Offer comparison blank (CSV)
• Preparation custody blank (CSV)
• Process map (MMD)
• Quantity bridge blank (CSV)
• Sample evaluation blank (CSV)
• Scoping and unit card blank (CSV)

That is the purpose of the full process: make each next step useful, authorized and specific enough that the following person can act on it. If you have existing footage and want to establish the first step, appraise your video with Rightmark. Start with the description and rights questions; the initial appraisal requires no files or system access.