# AI Avatar Personalised Video > Use when AI Avatar Personalised Video is needed to produce a AI avatar personalised video deliverable for social-media or digital-marketing work; use `ai-readiness-diagnostic` when its narrower outcome is requested. - Skill: `peterbamuhigire/ai-avatar-personalised-video` (Agent Skill) - Install (CLI): `npx skillmds@latest add peterbamuhigire/ai-avatar-personalised-video` - Raw SKILL.md: https://api.skillmd.com/api/skills/peterbamuhigire/ai-avatar-personalised-video/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Marketing & Growth - Author: peterbamuhigire (https://skillmd.com/u/peterbamuhigire) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/peterbamuhigire/ai-avatar-personalised-video --- # AI Avatar Personalised Video ## Use When - Use this skill when the requested outcome is specifically a **AI avatar personalised video deliverable** and the supplied brief falls within ai avatar personalised video. ## Do Not Use When - Use `ai-readiness-diagnostic` when its narrower output is the real deliverable; do not use this skill as a generic substitute. - Do not use it to publish, send, spend, alter a live account, or make unsupported legal, platform, performance, or certification claims. ## Required Inputs | Artefact | Source/provider | Required? | If absent | |---|---|---:|---| | AI marketing use-case brief, intended human control point and success measure | Requester or approved brief | Yes | Stop and request the missing decision context. | | Brand voice, offer facts, constraints and approvals | Client source pack or authorised owner | Conditional | State assumptions; do not invent names, prices, results or approvals. | | Performance, platform or research evidence used for claims | Traceable export, URL, document or named source | Conditional | Draft the narrowest reviewable version and flag the missing evidence. | ## Capability and Permission Boundaries Drafting is permitted within the supplied brief. Publishing, sending, spending, changing live accounts, or claiming certification requires separate explicit authority. Minimum capabilities are read access to supplied files and search across the authorised evidence set. Use only the files, tools, accounts and evidence made available for the engagement, expose every unassessed check, and obtain explicit authority before any mutation. ## Degraded Mode Fallback: if files, network access, platform data, language review or production tools are unavailable, return the narrowest useful qualified AI avatar personalised video deliverable; mark unavailable checks `not assessed` and never convert them into a pass. ## Decision Rules | Choice | Action | Failure or risk avoided | |---|---|---| | Data readiness, AI maturity and risk support the proposed operating level | Choose the lowest viable automation level and define its human approval gate. | Automating an unsafe or unevaluable marketing process. | | A required fact or approval is missing | Stop that claim or action; request it or use an explicit placeholder. | Fabricated facts, implied consent or unauthorised publication. | | Evidence is partial but a useful draft is possible | Deliver a qualified draft with gaps and the next verification step. | Treating an unassessed requirement as passed. | ## Workflow 1. Confirm the exact AI avatar personalised video deliverable, consumer, market, channel and approval boundary; route to `ai-readiness-diagnostic` if it is the closer match. 2. Inventory supplied facts, source provenance, constraints and missing inputs; stop if the objective, audience or authority is unknowable. 3. Select the domain method and record the material decision behind it before drafting. 4. Produce the smallest complete AI avatar personalised video deliverable; keep facts traceable and placeholders visibly unresolved. 5. Test the result against the decision table, domain quality criteria and anti-slop gate; recover by narrowing or qualifying unsupported portions. 6. Deliver the artefact with evidence, assumptions, unassessed checks and the next approval or verification step. ## Outputs | Artefact | Consumer | Observable acceptance condition | |---|---|---| | Ai avatar personalised video deliverable | Requester, client reviewer or delivery team | The AI avatar personalised video deliverable addresses the named audience and objective, records assumptions, and passes the skill's domain checks without invented facts. | | Decision and gap note | Approver or next workflow | Names the chosen route, evidence used, unresolved inputs and any action requiring authority. | ## Evidence Produced | Evidence | Format | Acceptance condition | |---|---|---| | Source/assumption register and completed release checklist | Inline table, checklist or linked source note | Every material claim, decision and unavailable check is traceable. | ## Quality Standards - Preserve the domain guidance and East African market context below; replace it only when the requester names another market. - Use British English unless the target language or market requires otherwise, and verify names, figures, quotations and platform rules before use. - Make the key choice visible, cover failure and edge cases, and keep the result ready for its named consumer. - Run the repository's `anti-ai-slop` ship gate; a blocking factual, cultural, safety or permission defect stops release. ## Anti-Patterns - Writing before the objective and audience are known. **Fix:** stop and obtain the missing brief fields. - Reusing a neighbouring skill's template because the headings look similar. **Fix:** route by the requested AI avatar personalised video deliverable, not vocabulary overlap. - Adding a price, result, quotation, platform limit or cultural claim without a traceable source. **Fix:** verify it or qualify/remove it. - Treating missing access, evidence or native-language review as approval. **Fix:** mark the check `not assessed` and narrow the result. - Publishing, sending, spending or changing a live account from drafting authority alone. **Fix:** obtain explicit action-specific authority and retain the approval record. ## References - [ai-readiness-diagnostic](../ai-readiness-diagnostic/SKILL.md) is the nearest routing comparison for this skill. - [Repository agent guide](../../../AGENTS.md) defines the engine-wide market, safety and anti-slop gates. ## Purpose Produce operational plans and production-ready scripts for AI avatar video deployments using Synthesia, HeyGen, Tavus, D-ID, or Elai.io. Output covers tool selection, avatar design, script writing, distribution, and disclosure compliance. Every deliverable is sized for the East African market by default. > Roth, H. and neuroflash Team (2024/2025) *AI Strategy 2025 for Marketing > Teams*. neuroflash. [Statistics on video outreach performance cited below] ## Required Input Before generating any deliverable, ask for: 1. **Client business name** and industry 2. **Country and city** (default: Uganda / Kampala) 3. **Primary goal** — outreach at scale, client communications, social content, or onboarding/training 4. **Volume required** — estimated number of videos per month and whether personalisation per recipient is needed 5. **Existing human presenter** — is there a real person whose voice and likeness the avatar should represent, or will a stock avatar be used? 6. **Budget tier** — indicative monthly budget in USD (see platform pricing below) ## Section 1 — Tool Selection Guide ### Platform Comparison | Platform | Best for | Avatar creation | Languages | Free tier | |---|---|---|---|---| | Synthesia | Corporate training, branded explainers, polished templates | 230+ stock avatars; custom avatar from upload | 140+ | No (14-day trial) | | HeyGen | B2B video outreach, language localisation, scalable personalisation | Instant from 60-second clip | 40+ | Yes (1 free video/month) | | Tavus | 1:1 personalised video at scale (unique video per recipient) | Clone from 5-minute recording | English primary | No | | D-ID | Talking-head from photo; fast turnaround; no minimum commitment | Animated photo or uploaded image | 119 | Yes (20 credits free) | | Elai.io | Volume production at lower cost; best value for EA price point | Custom avatar from video upload | 75+ | Yes (1 free video/month) | ### Selection Rules Apply these rules to select the right platform before starting production: - **Outreach volume > 100 personalised videos/month:** Use Tavus. Each recipient receives a unique video with their name, company, and a personalised line. - **Polished evergreen content (training, onboarding, explainers):** Use Synthesia. Templates are the strongest in the market; avatar quality is broadcast-standard. - **Language localisation required (Swahili, French, Luganda voiceover):** Use HeyGen. Its translation feature re-lips the avatar in the target language from a single English recording. - **Quick turnaround from a single photo, minimum budget:** Use D-ID. Ideal for one-off client communications or prototype testing. - **Volume production on a constrained budget (under USD 50/month):** Use Elai.io. The price-to-output ratio is the strongest for high-frequency social content. ### EA Pricing Reference (March 2026) | Platform | Entry paid plan | Approx. UGX/month | |---|---|---| | Synthesia | USD 29/month (Starter — 10 min video/month) | ~UGX 107,000 | | HeyGen | USD 29/month (Creator — 15 credits) | ~UGX 107,000 | | Tavus | USD 150/month (Starter) | ~UGX 555,000 | | D-ID | USD 9/month (Lite — 10 min) | ~UGX 33,000 | | Elai.io | USD 23/month (Basic — 15 min) | ~UGX 85,000 | Flag: all platforms require payment by international card or PayPal. MTN Mobile Money payment is not directly supported; use Payoneer or Wise to fund an international card from a Mobile Money wallet. ## Section 2 — Avatar Design to Brand Specification ### Wardrobe and Visual Identity - Specify wardrobe colour in the design brief using the client's primary brand colour hex code; most platforms allow uploading reference images. - Avoid busy patterns or small checks — they compress poorly in MP4 output. - Logo placement: Synthesia and HeyGen support branded backgrounds with logo overlays; for platforms that do not, use a branded frame in Canva applied in post-production. - Professional register must match the sector: formal attire for financial services and legal; smart-casual for technology and creative sectors; branded casual (polo shirt with embroidery) for retail and FMCG. ### Background - **Branded background:** Use the client's office, a branded set mockup, or a solid brand-colour background. Synthesia, HeyGen, and Elai.io all support custom background image uploads. - **Neutral background:** A plain white, grey, or dark studio background works for B2B outreach where the message must carry the video, not the visual. - D-ID does not support custom backgrounds natively; apply background via CapCut or Canva's video editor before distribution. ### Voice: ElevenLabs Integration Where the client has a human presenter whose voice should match the avatar: 1. Record the presenter reading a 3–5 minute voice sample (clear, neutral pace, no background noise, on a phone in a quiet room is sufficient). 2. Upload to ElevenLabs Voice Cloning (Starter plan: USD 5/month). 3. Export the cloned voice as an MP3 and import into Synthesia, HeyGen, or Elai.io via their "custom voice" upload function. 4. Test with a 30-second sample video before full production. 5. Specify pace in the ElevenLabs settings: 0.9x speed is optimal for East African professional audiences; 1.0x can read as rushed. ### Approval Protocol Obtain written client sign-off (email confirmation is sufficient) on: 1. Avatar appearance (screenshot from the platform's preview) 2. Voice sample (30-second MP3) 3. Disclosure language (see Section 6) 4. One full pilot video before any volume production begins Do not proceed to volume production without approval at each stage. ## Section 3 — Script Writing for AI Avatars Script quality is the primary determinant of avatar video quality. AI avatar delivery requires a fundamentally different structure to live video. ### Core Rules - **Sentence length:** 15 words maximum per sentence. Break longer sentences at natural breath points. - **No compressed contractions:** Avoid "I'd've", "we'd've", "they'd've" — write "I would have", "we would have". Standard contractions (I'm, we're, it's) are acceptable. - **No idioms:** Avoid "hit the ground running", "move the needle", "circle back". Write the plain meaning: "start immediately", "improve results", "follow up". - **Pause markers:** Insert `[PAUSE]` wherever a human speaker would breathe, hesitate, or let a point land. - **Emotional register instructions:** Open with a register instruction in parentheses — e.g. `(warm, direct)` or `(confident, measured)`. Insert a new instruction before any tonal shift — e.g. `(slight urgency)` before the CTA. - **EA register:** Avoid English idioms that do not translate to the East African professional context. Write in the tone expected in an EA boardroom: direct, respectful, and specific. ### Length Limits - **Outreach video (cold or warm B2B):** 60–90 seconds maximum. Anything longer loses the recipient before the CTA. - **Client communications (e.g. monthly report summary):** 90–120 seconds. - **Onboarding or training explainer:** Up to 3 minutes. Longer content must be broken into chapters (each chapter = a separate video file). - **Social content (Reels, TikTok, Facebook short video):** 30–60 seconds. ### Script Template — B2B Outreach (60–90 seconds) ``` (warm, direct) Hello [FIRST NAME]. [PAUSE] My name is [PRESENTER NAME] from [COMPANY]. I noticed [ONE SPECIFIC, RESEARCHED OBSERVATION ABOUT THEIR BUSINESS]. [PAUSE] We help [TARGET AUDIENCE DESCRIPTION] to [PRIMARY OUTCOME]. [PAUSE] One of our clients, [CLIENT REFERENCE], achieved [SPECIFIC RESULT] in [TIMEFRAME]. (slight urgency) I would like to show you how we could do the same for [THEIR COMPANY NAME]. [PAUSE] Reply to this message and I will send you a 20-minute briefing. [PAUSE] Thank you. ``` ### Script Template — Monthly Report Summary (90–120 seconds) ``` (professional, warm) Hello [CLIENT NAME] team. [PAUSE] Here is your [MONTH] performance summary. [PAUSE] This month, your top-performing content was [POST TITLE OR DESCRIPTION]. It reached [REACH FIGURE] people and generated [ENGAGEMENT FIGURE] interactions. [PAUSE] Your follower growth was [FIGURE] — [UP/DOWN] [PERCENTAGE] from last month. (direct, confident) Three priorities for [NEXT MONTH]: [PRIORITY 1]. [PAUSE] [PRIORITY 2]. [PAUSE] And [PRIORITY 3]. [PAUSE] The full report is attached. [PAUSE] I am available Thursday if you would like to discuss any of this. Thank you. ``` ## Section 4 — Use Cases with EA Application ### B2B Video Outreach Personalised video emails achieve 75% open rates and up to 40% response rates, compared to 21% average open rates for text-only emails (Roth and neuroflash Team, 2024/2025). - **Tool:** HeyGen (for variable personalisation with the recipient's name and company inserted via CSV) or Tavus (for fully unique video per recipient). - **Personalisation variables:** First name, company name, one researched observation about their business (sourced from their LinkedIn or website). - **Distribution:** LinkedIn InMail with video thumbnail embedded + follow-up email with GIF preview. Do not attach the raw MP4; host on Loom, Vimeo, or the platform's share link and embed the thumbnail. - **EA application:** B2B outreach in Uganda/Kenya is relationship-led. Lead with a warm, specific observation before any product or service mention. A generic "I saw your company" opener will not perform in this market. ### Client Communications - Monthly performance report as a 2-minute avatar video from the account manager. Replaces or supplements the PDF report. - Use the account manager's own avatar (HeyGen custom avatar from a 60-second clip) so the client sees a familiar face. - Distribute via WhatsApp (under 10MB MP4) and email (thumbnail link). - EA application: personalised video report increases perceived value of the retainer. Clients who see a face presenting results are less likely to question the agency fee. ### Onboarding - Welcome video for new clients: 90-second avatar video from the managing director, welcoming the client and confirming the first three steps. - Platform training introduction: a 3-minute explainer on how to read their monthly dashboard or how to request content amendments. - Use Synthesia for evergreen onboarding content — the polished templates are appropriate for a first impression. ### Social Content - Short branded videos for Instagram Reels (30 seconds), TikTok (30–45 seconds), and Facebook (45–60 seconds). - The avatar handles the talking-head element; pair with B-roll footage or branded graphics in CapCut or Canva. - Production time: under 30 minutes per video once the script and avatar are approved. - EA application: Ugandan audiences respond strongly to a recognisable, consistent presenter. A recurring avatar character builds familiarity over time in the same way a brand spokesperson does on television. ## Section 5 — Distribution Protocol ### LinkedIn - Post the video directly to the LinkedIn feed (native video outperforms links in the algorithm). - For outreach: send as an InMail with a thumbnail image embedded in the message body and the video URL beneath it. - Add captions; LinkedIn auto-generates them — edit for accuracy before posting. ### Email - Do not embed the MP4 file directly. Host the video and use a static thumbnail with a play-button overlay as the email image. - A GIF preview of the video's first 3 seconds increases click-through rate by 26% (Roth and neuroflash Team, 2024/2025). - Subject line format: "Video for [FIRST NAME]" or "[FIRST NAME] — 90 seconds on [TOPIC]" consistently outperforms generic subject lines. ### WhatsApp (EA Primary Channel) - Compress the MP4 to under 10MB before sending (use HandBrake, free, or CapCut's export settings). - Send directly to known contacts via WhatsApp Business. - For broadcast to segments: use WhatsApp Business broadcast lists (maximum 256 contacts per list). Do not broadcast unsolicited video; send only to contacts who have opted in to communications. - A voice note teaser (15 seconds, human voice) sent before the video link increases video opens in the EA market — the voice note signals a genuine personal message rather than a mass broadcast. ## Section 6 — Disclosure Requirements ### Legal Obligation Under EU AI Act Article 4 (effective August 2026) and emerging global norms, any video using an AI avatar that is not a real person must be disclosed as AI-generated. This requirement applies to all clients regardless of their country of operation, because: - Content may be received by EU-based contacts or distributed on EU-regulated platforms. - Non-disclosure is a reputational risk in any market; disclosure done well is a brand asset. ### Standard Disclosure Language Include this line in every AI avatar video communication: > "This message was delivered by an AI avatar on behalf of [FULL NAME], > [TITLE] at [COMPANY NAME]." Place the disclosure: - As the final sentence of the video script (spoken by the avatar) - In the email footer or LinkedIn message body beneath the video link - In the video description on any platform that supports it ### Custom Avatar Disclosure When the avatar uses a real person's likeness and voice (cloned via ElevenLabs or a platform's custom avatar tool), add: > "This is an AI-generated video using a digital likeness of [NAME], created > with their consent." ### Client Approval Obtain written client approval of the disclosure approach before any video is sent to a third party. Document the approval in the client's folder. ## Quality Criteria - The correct platform is selected using the selection rules in Section 1; the rationale is stated in one sentence referencing the client's primary goal and volume requirement - The avatar design brief specifies wardrobe, background, voice, and pace with reference to the client's brand standards and sector register - Every script follows the 15-word sentence limit, contains at least two `[PAUSE]` markers, and opens with a parenthetical register instruction - Scripts are timed before production; outreach scripts are 90 seconds or under when read at the specified pace - Distribution is specified per channel with file size limits and thumbnail instructions for email and WhatsApp - Disclosure language is included in every deliverable and written client approval is required before volume production - EA market context is reflected throughout: WhatsApp distribution, payment method feasibility, EA professional register, and relationship-led outreach norms - Volume production does not begin until the pilot video has been approved in writing ## Related Skills - **`biz-dev-video-outreach`** — use for the full video outreach sequence (messaging strategy, follow-up cadence, CRM tracking) that AI avatar videos feed into - **`brand-voice-ai-training`** — use to establish the brand voice parameters before scripting avatars; the brand voice brief feeds directly into the register instructions in each script - **`playbook-content-production`** — use for the full content production workflow that AI avatar social videos sit within ## References - Roth, H. and neuroflash Team (2024/2025) *AI Strategy 2025 for Marketing Teams*. neuroflash. [Video outreach performance statistics; distribution CTR data] - Chaffey, D. (2024) *Digital Marketing: Strategy, Implementation and Practice*. Pearson. [Channel strategy context for distribution sequencing]