Correction, 2026-10-02: the union section rested on a law firm's summary and a newspaper reprint. It now quotes SAG-AFTRA's own documents, which differ in two places: the 48-hour rule is about seeking consent to create a digital replica (24 hours when the performer is engaged later), not about describing the use 48 hours ahead, and the payment floor is "no less than 1.5 session fees" per commercial, not 1.5 times the scale session fee. We removed the paragraph on the June 2026 TV/Theatrical agreement, because we could not read SAG-AFTRA's own page on it today, and added the contract's Synthetic Performer terms, which a union signatory should check before using a synthetic voice. We had not said that 8frame has no voice cloning, translation or dubbing tool, that Eleven v3 is the default TTS model, that the canvas Montage node drops sound so the read goes on the Video Editor timeline, or that a speaking face needs a lip-sync clip of up to 35 seconds. We also rewrote claims we had not measured (that the first pass gets pacing wrong, that punctuation and respelling reliably fix a read, that mispronunciation is the most common tell, that translations often run longer, that stock voices are the fastest class) as things to check.
Correction, 2026-10-01: an earlier version quoted voiceover talent rates, AI voice subscription prices and multilingual budgets we could not trace to a source, gave the 8frame TTS price without saying it is charged per started 1,000 characters, and named Replica Studios, Narrativ and Ethovox as current SAG-AFTRA-approved platforms that let members license their own voices. We could not confirm that those arrangements work that way or are still in force, so they are removed. The union section now cites the 2025 Commercials Contracts, and the cost table uses 8frame prices and the contract terms.
A professional voiceover for a commercial means booking talent and studio time, then paying for usage: union commercial work is paid as a session fee plus use and holding fees. AI voiceover turns that into a paste-the-script-and-generate step you can repeat until the read is right, paying per generation. The tradeoff is that "usable" depends heavily on the class of voice you use and the QA you do afterward. This guide covers the three classes of AI voice, the union rules you have to respect in 2026, how to cast and direct a synthetic read, multilingual VO, and the checklist that keeps a synthetic read from sounding synthetic.
8frame generates the read too: ElevenLabs TTS runs on the canvas at 13 credits per started 1,000 characters (a 1,200-character script bills 26), up to 5,000 characters per generation, on Eleven v3 (the default) or Multilingual v2 at the same price. It reads with a list of preset ElevenLabs voices, and 8frame has no voice cloning and no dubbing tool, so a read made there is the stock-voice class below; the voice and the footage come from the same credit pool. You can still bring a read from an outside voice tool if you prefer its library, then cut it against the footage you built on 8frame. The reason it belongs in your ad workflow is the same volume logic that drives AI performance marketing: when a revision costs a regeneration instead of a re-book, you can test five reads of a hook instead of defending the one you paid for.
The three classes of AI voice
Not all synthetic voice is the same thing legally or qualitatively. Pick the class before you pick the tool.
1. Text-to-speech (TTS) stock voices. A library of pre-built synthetic voices the vendor owns and licenses to you. No specific human is being replicated, which makes this the safest class for consent. Whether a given stock voice is good enough for a produced spot is a judgment to make read by read against your footage; controlled reads (announcer, explainer, retail) are the easiest place to start. If you produce under a SAG-AFTRA contract, also read the Synthetic Performer terms below.
2. Cloned voices. A voice model trained on a specific person's recordings, used to generate new speech in that person's voice. 8frame has no voice cloning, so this class means another tool. This is where the rules bite. Cloning a real person, talent, a founder, anyone, needs that person's explicit consent, and if they are union talent, the contract's digital-replica terms. Cloning without consent is exactly what those terms are written to stop.
3. Licensed synthetic voices. Voices built from the recordings of performers who agreed to it and are paid for its use. The appeal is a distinctive, human-grounded voice with a rights chain behind it. These arrangements come and go, so ask the vendor in writing which agreement covers the voice today, whether it is still in force, and what uses it allows, and keep that answer in your rights file.
The SAG-AFTRA rules you have to respect
If you are producing commercials in the US with union talent, treat the union framework as a hard constraint, not a nice-to-have. SAG-AFTRA members ratified the 2025 Commercials Contract and the 2025 Audio Commercials Contract on May 21, 2025, by 96.90% to 3.10%; both run three years from April 1, 2025, and the audio contract's A.I. section points to the same language. In SAG-AFTRA's own summary of the new provisions:
- Digital replica: "a computer program made in whole or in part using the voice, image and/or performance of a performer that can independently generate new performances". A voice clone built from a performer's recordings fits that definition.
- Notice: the producer must use commercially reasonable efforts to obtain consent to create a digital replica at least 48 hours in advance, or at least 24 hours when the performer is engaged less than 48 hours ahead.
- Consent: the performer gets "a reasonably specific description of intended use", and consent "must be clear and conspicuous", given in the Digital Replica Rider or in "a separate, signed writing". A use beyond that description needs new consent.
- Pay: "no less than 1.5 session fees for each commercial" that contains a performance generated by the replica (unless the performer's session pay already covered that performance), plus use and holding fees.
- Synthetic performers: a digitally created asset that gives "the clear impression" of a human performer who is not identifiable, is "Not voiced by a human performer" and is not a digital replica. A signatory producer "shall not use a Synthetic Performer primarily or materially for reasons of economy or economic advantage in the engagement of performers", and creating one by prompting with an identified performer's name, photograph, image or recording needs that performer's consent and at least the minimum session fee.
What this means when you sit down to make an ad:
- Stock TTS voices owned by the vendor copy no performer, so there is no replica consent to get. If your company or agency is a SAG-AFTRA signatory, ask the union or your counsel whether a synthetic voice in your spot counts as a Synthetic Performer before you rely on it.
- Cloning a real person requires written consent, and if they are union talent, the replica terms above. No exceptions, no "it's just for a test."
- Licensed synthetic voices are only as clean as the agreement behind them: get the rights chain in writing before the spot runs.
The reputational risk is as real as the legal one: a campaign that uses a voice cloned without consent becomes a story about that clone.
The 4-step workflow
1. Cast the voice
Write the delivery you want the way a director would: age, tone, pace, energy, accent, and emotional read. Some voice tools can design a voice from a description like this; on 8frame you pick from the preset ElevenLabs voices, so the direction becomes your casting brief when you audition them. Instead of a vague "friendly female voice," specify the delivery you want.
An example direction for a DTC wellness spot:
Warm, grounded female voice, late 30s, unhurried pace. Conversational,
like she is telling a friend something she believes. Slight downward
inflection at line ends, not upselling. No radio-announcer energy.
Medium-low pitch. American, neutral.
Generate the same script with two or three voices and audition them against the footage; a script under 1,000 characters is 13 credits per voice. Voice-to-picture fit decides it: a great read that does not match the on-screen energy is the wrong read.
2. Direct the read through the text
If a line comes back with the wrong pacing or emphasis, fix it deliberately rather than regenerating blindly. On 8frame the lever is the script itself: the TTS node takes the text, the voice, and the model, nothing else. Try punctuation to shape the pauses, a full stop before the product name for a beat, shorter sentences to slow the CTA, and a phonetic respelling for any word the model mispronounces (brand names and unusual terms are the ones to check first). Listen to each retake; each one is another charge of 13 credits per started 1,000 characters.
3. Generate the multilingual versions
This is where AI voice earns its place outright. Once your English read is approved, generate the Spanish, German, French, and Portuguese versions from translated scripts with the same preset voice, without re-booking native talent per market. ElevenLabs lists all four languages for both Eleven v3 ("70+ languages") and Multilingual v2 ("29 languages"). 8frame has no translation or dubbing tool, so you supply each translated script. Each language is its own charge of 13 credits per started 1,000 characters, and translated copy can run longer than the English, so a script just under 1,000 characters in English can bill 26 in German. Pair this with a proper localization pass so the copy is transcreated, not machine-translated, and keep a native speaker in the QA loop for each language. See the AI localization workflow for the full pipeline.
If the spot has no speaking face and no on-screen text, the picture travels unchanged and only the audio swaps, so one video cut serves every market. If a presenter speaks on screen, each language needs its own lip-sync clip: 8frame's lip-sync tools make a new clip from a still portrait and the audio, up to 35 seconds (prices in AI lip sync price by model), and none of them re-syncs the mouth on footage you already have.
4. Mix against the 8frame cut and QA
Put the approved read and your AI music bed on the 8frame Video Editor timeline, which has sound layers, or in your own editor, over the cut. Do not join talking clips with the canvas Montage node: its merged file has no sound, so voice and music go on the timeline. Give the read and the bed their own sound layers, set each layer's volume so the voice sits above the music, and confirm the read lands on the picture beats. Then run the QA checklist below before it ships.
QA checklist for synthetic reads
Run every read against this before it goes to a paid placement:
- Mispronunciations: brand names, product names, place names, numbers, and any acronym.
- Breath and pacing: does it breathe like a person, or run lines together unnaturally? Add pauses where a human would.
- Emphasis: is the stress on the right word in the sentence? Wrong emphasis is subtle but reads as "off."
- Consistency: across a multi-line script, does the voice keep the same character, or drift in tone between generations? Generate the full script in one pass where possible (up to 5,000 characters on 8frame).
- Emotional match: does the read's energy match what is on screen at that moment?
- Rights: is this a stock voice, a consented clone, or a licensed synthetic voice? Documented before it ships.
Cost math
| Path | What you pay | Revisions |
|---|---|---|
| Stock TTS on 8frame (ElevenLabs) | 13 credits per started 1,000 characters; a 30-second script (about 75 words) is one charge of 13, about $0.13 at pack rate | Each new take is another charge |
| Licensed synthetic voice | The vendor's plan plus its per-use terms | Per the vendor's terms |
| Union VO talent | Session fee plus use and holding fees | More paid session time for pickups |
| Digital replica of a union performer | Clear, written consent and no less than 1.5 session fees per commercial, plus use and holding fees | Uses beyond the agreed description need new consent |
Multilingual versions show the difference most plainly. Five languages of stock TTS from an approved script are five charges of 13 or 26 credits each (65 to 130 credits, $0.65 to $1.30 at pack rate), against a booked session per language.
FAQ
Is it legal to use AI voiceover in a commercial?
Yes, when you use the right class of voice. Stock TTS voices owned and licensed by the vendor copy no performer and are the lowest-friction option for consent. Cloning a specific real person's voice requires that person's explicit written consent and, if they are union talent, the replica terms of the SAG-AFTRA Commercials Contracts: clear, written consent to a reasonably specific use, and no less than 1.5 session fees per commercial plus use and holding fees. If you are a union signatory, check the contract's Synthetic Performer terms before you use a stock synthetic voice. Licensed synthetic voices are fine when the vendor can show you the agreement behind the voice. The one thing you cannot do is clone a real voice without consent.
Can AI voice really pass for a human read in a produced spot?
Judge it read by read against your footage rather than trusting a general claim. Controlled reads (announcer, explainer, retail, corporate) are the easiest to get right with direction and QA. Highly emotional or performance-driven delivery, a read that has to carry vulnerability, comic timing, or a character, is where a human performance is still the safer choice. Use AI for the workhorse reads and the multilingual long tail, and keep human talent for the performance-led hero spot.
How do I do multilingual voiceover without booking talent in every market?
Approve your primary-language read, then generate each additional market's version from a properly transcreated (not machine-translated) script using a matching AI voice. On 8frame you supply the translated script, since there is no translation or dubbing tool. Keep a native speaker in the QA loop for each language to catch pronunciation and tone issues, since that is exactly where machine reads slip. Without a speaking face or on-screen text, the video stays the same across all markets and only the audio swaps. See the AI localization workflow for the end-to-end process.
Can 8frame clone my voice or dub an existing ad?
No. 8frame's ElevenLabs TTS reads with preset voices; there is no voice cloning, translation or dubbing tool. Its lip-sync tools make a new clip from a still portrait and an audio track, up to 35 seconds, rather than re-syncing footage you already have.
Sources
- 8frame ElevenLabs TTS price (13 credits per started 1,000 characters, Eleven v3 by default and Multilingual v2 at the same price), 5,000-character limit, the node's inputs (text, voice, model) and preset voice list: the 8frame app, checked 2026-10-02
- No voice cloning, translation or dubbing tool; lip-sync tools make a new clip from a portrait and audio, up to 35 seconds; Montage output has no sound; Video Editor timeline with sound layers: the 8frame app, checked 2026-10-02. Volume per sound layer: the 8frame app, checked 2026-10-02
- 2025 Commercials Contract and Audio Commercials Contract ratified May 21, 2025, by 96.90% to 3.10%: SAG-AFTRA, Members Approve 2025 Commercials Contracts Agreement (sagaftra.org, as archived on 2025-06-02), read 2026-10-02
- Digital Replica, notice, consent, payment and Synthetic Performer terms, three-year term from April 1, 2025, and the audio contract's pointer to the same A.I. language: SAG-AFTRA, Summary of New Provisions, 2025 SAG-AFTRA Commercials Contract, section 42, and the Audio Commercials Contract summary in the same file, section 17 (sagaftra.org PDF, as archived on 2025-05-13), read 2026-10-02. sagaftra.org returned HTTP 403 to our direct requests on 2026-10-02, so both quotes come from SAG-AFTRA's own pages in the Internet Archive
- Eleven v3 and Multilingual v2 languages: ElevenLabs, Models, checked 2026-10-02
- Credit value ($0.01 per credit at pack rate, $10 for 1,000 credits): 8frame pricing, checked 2026-10-02
The read is only half the spot. Build the picture on the same canvas (Veo 3.1, Kling, Seedance and the other video models), then drop your AI voiceover and music bed over a cut you finished the same day.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.