Radio Drama and Voice Cloning: A Casting Guide for Small Production Teams

Build the casting plan around dramatic function

Small audio drama companies often work with limited rehearsal time, a modest cast and several overlapping production responsibilities. Voice cloning may appear to solve a practical casting problem, particularly when a recurring character needs a short additional line after recording has finished. However, the first decision should still be artistic: what does the role require, and what kind of performance will make the scene work?

A production with four principal characters and several incidental speaking roles might use a human cast for the main performances while considering synthetic production for narrowly defined supplementary material. A background announcement, for example, has different dramatic requirements from a character delivering a confession. The distinction should be made by the director and producer, not by the availability of a voice-generation feature.

Prepare a role breakdown before approaching performers. Identify speaking duration, emotional range, recurrence, improvisation requirements and the possibility of future episodes. This gives the company a clearer understanding of where a synthetic voice might be technically useful and where a performer brings essential interpretative value.

• Principal characters: prioritise performance, chemistry and continuity.
• Recurring secondary characters: consider availability across the planned series.
• Incidental voices: establish whether a recorded human performance is practical.
• Additional lines: consider synthetic production only where permission is explicit.

Consent must describe the actual production

A voice is part of a performer's professional identity. Permission to record dialogue for a radio play does not automatically grant permission to train a voice model, generate new dialogue or reuse a performance in later episodes. These are different activities with different implications. A production company should describe the proposed cloning process before recording begins and give performers a meaningful opportunity to accept or decline it.

The permission should identify the material that may be used to create a model, the character or project involved, the intended output, the people who can access the system and the period during which permission applies. It should also address whether generated lines can be altered, whether the model can be used in promotional recordings and what happens when the production ends.

A performer should not have to accept voice cloning as an unexplained condition of being cast. Where a production can use an alternative performance method, that option should be discussed. Agreements about payment, attribution, model retention and withdrawal require careful drafting, and the company should obtain appropriate professional advice for its jurisdiction.

Serialised fiction needs a continuity system

Serialised audio fiction creates a specific challenge: characters and their relationships develop over time. A voice heard in episode one may return several months later with a changed emotional position or a different understanding of events. A synthetic voice that preserves superficial vocal characteristics cannot independently maintain the character's dramatic history. That continuity belongs to the writing and directing process.

A series bible should document each character's established voice, speech habits, relationships, pronunciation and significant changes in emotional behaviour. It should also distinguish deliberate development from accidental inconsistency. If a character becomes more guarded after a betrayal, the change should be an intentional performance decision rather than an unexplained variation in delivery.

Episode structure also affects the production method. A short instalment ending on a cliffhanger may require precise timing and a carefully controlled final line. A longer episode with multiple scenes may place more weight on sustained character relationships. The format should inform casting and recording decisions before the production team commits to a particular workflow.

• Maintain a character voice and pronunciation guide.
• Record established character relationships and emotional developments.
• Track recurring names, locations and story-specific terminology.
• Identify which episode moments require direct performer approval.

Where radio drama AI tools can help

Radio drama AI tools cover several different production functions, and they should not be treated as one interchangeable category. Speech recognition can assist with dialogue transcription. Noise reduction may help address an unwanted background sound. Synthetic speech can support an authorised additional line, while editing and audio restoration tools can assist with technical preparation. Each task has different creative and contractual implications.

For a small production team, transcription can be particularly useful when several scenes have been recorded across different sessions. A searchable transcript helps locate a line, compare versions and organise revisions. Speech separation may also assist with difficult recordings, although overlapping dialogue and complex room sound can remain challenging. The original multitrack recordings should remain available for editorial decisions.

Voice cloning is more consequential because it can generate speech associated with a recognisable performer. The company should separate ordinary editing permissions from model creation and synthetic performance permissions. A licence for one production function should not quietly become permission for another. The producer should document which tools are used, what material enters them and who approves the resulting audio.

A small-team workflow needs named responsibilities

A lean production can still have clear departments, even when one person performs several jobs. Assign responsibility for casting, performer agreements, recording management, synthetic production, sound design and final editorial approval. If the director also operates the recording system, another person should have a clear route to raise concerns about generated dialogue or inappropriate use of a performer's material.

A practical sequence begins with the script and casting breakdown. The company then agrees the recording and cloning permissions, records the human performances, completes the edit and identifies any genuinely necessary additional lines. Any authorised synthetic material should be reviewed in the context of the complete scene. The final mix should be checked for continuity, intelligibility and consistency with the character's established performance.

File management is part of this process. Keep source recordings, approved edits, generated material and final masters in clearly separated project folders. Access to voice models should be restricted to the people who need it. Establish a retention and deletion procedure that reflects the production agreement and the relevant service terms.

• Producer: permissions, budgets and production records.
• Director: character interpretation and performance approval.
• Sound editor: dialogue preparation, continuity and final mix.
• Project administrator: access control and agreed retention procedures.

Credit and payment should follow the contribution

Credits should explain meaningful creative contributions instead of allowing synthetic production to blur the identity of the people involved. A performer who records the source material for an authorised voice model has contributed something distinct from a performer who delivers every final line. Both contributions can be recognised accurately. The production should avoid suggesting that generated speech was performed live by the person whose voice it resembles.

Payment arrangements need to reflect the agreed scope. A recording fee, a voice-model creation fee and compensation for subsequent generated uses are different possible elements of an agreement. There is no universal rate that applies to every independent audio drama. The parties should establish the basis of payment before production and address whether additional episodes, promotional clips or adaptations require new permission or compensation.

A small company may be unable to support an extensive legal and administrative department, but that is a reason to simplify the production plan, not to leave essential rights undefined. If the terms are too complex to administer reliably, a conventional recording or a different casting approach may be the more workable choice.

Make the casting rules part of the series bible

The most useful time to establish synthetic voice rules is before the first episode is recorded. The series bible can state which characters are performed by human actors, whether any authorised synthetic production is anticipated and who can approve changes. It can also explain how a recast, a performer departure or an unexpected additional episode will be handled.

For example, a six-episode mystery might have three principal performers and two recurring supporting roles. The production could decide that all principal dialogue remains human-performed, while any synthetic supplementary speech requires separate written approval from the relevant performer. That rule is simple enough to administer and specific enough to prevent informal expansion of permission.

Before the first casting session, prepare a one-page production policy covering performance methods, consent, credits, payment principles, access and retention. Circulate it alongside the role breakdown and have the relevant agreements reviewed by a qualified professional. The policy should be updated whenever the series format or intended uses change.

Questions readers ask