They are not waived for a large customer, an urgent deadline, or a persuasive argument about intent. Several are enforced as blocking positions in the rights ledger rather than as a review at the end.
01
Production we decline
- Political persuasion, electioneering, or content designed to influence a vote
- News, or anything presented as a factual account of real events
- Financial advice, investment claims, or passive-income promises
- Deceptive testimonials, fake reviews, or fabricated endorsements
- Celebrity likeness or cloned voice without verified permission
- Depictions of real people, actors or franchises in generated artwork
- Content targeting a private individual
- Sexual content, and any sexualised depiction of minors under any framing
- Harassment, hate, or content that demeans a protected group
- Method instruction for crime, self-harm or weapons
- Bulk content-farm or faceless-video production
- Child-directed production, during the initial compliance phase
02
The permission rules
Likeness and voice are where synthetic production is most easily misused, and the rules are correspondingly rigid.
- Likeness
- No real person is depicted without verified permission recorded as a grant with a territory, term and expiry. "They would not mind" is not a permission.
- Voice
- No voice is cloned without verified permission on the same terms. This applies to public figures exactly as it applies to anyone else.
- Brand and product
- A real brand appears only where the brand is the customer or has granted permission, with substantiated claims per market.
- Disclosure
- Synthetic-media disclosure obligations attach to the deliverable and are checked in the delivery package. Removing a required disclosure is a violation of this policy.
03
Work that gets additional review
Some launch genres are permitted with an extra gate rather than declined, because doing them carelessly is the problem rather than doing them at all.
- Mythology, folklore and devotional materialA reviewer with standing in the tradition depictedBlocks: Boarding of the affected episodes
- Crime, harm and revenge-adjacent depictionCreative approver, against an agreed treatmentBlocks: Full-quality rendering
- Intimacy treatmentCreative approver, recorded at greenlightBlocks: Boarding and rendering
04
How this is enforced
- 01At qualification
Several of these categories are named as customers we do not take, before a brief is written.
- 02At the greenlight brief
Intended use is a field, and it is approved by a named person.
- 03At rights clearance
Likeness, voice and source permissions are blocking positions in the ledger. Missing ones stop the work that depends on them.
- 04At continuity and compliance QA
Claims, prohibited contexts, disclosure and caption safety are checked in the same pass as visual continuity.
- 05At delivery acceptance
The package does not hand over with an unresolved blocking position.
If you believe a production made with Tosheo violates this policy, write to [email protected] with enough detail to identify it. We would rather hear about it than not.
05
Questions people ask
Can I make a series about a real historical figure?
It depends heavily on the figure, the depiction and whether anyone living is identifiable. It is a conversation before a greenlight rather than a checkbox, and there are cases where the answer is no.
What if I have permission for a likeness?
Then it is recorded as a grant with a territory, term, platform and expiry, and the expiry is a date the system holds. Verified permission is the requirement; the verification step is not something we skip because the request is confident.
Is crime drama allowed?
Thriller and revenge-adjacent crime are launch genres with an additional review on depictions of harm. What is not allowed is anything about an identifiable real person or case presented as fact, or anything that functions as method instruction.
Why decline political content entirely?
Because the failure mode is severe, the incentives around it are bad, and a production company in bootstrapped validation has no business being a factor in an election. It is a category refusal rather than a judgement per request.
Does this apply to branded work too?
Yes, with additions: claims must be substantiated for the market they run in, prohibited contexts are brand canon and checked in QA, and spokesperson permissions are territory-bound grants like any other.