{"slug": "prepare-and-test-famulor-announcements", "title": "Prepare and Test Famulor Announcements", "summary": "Famulor's September 17, 2026 update lets teams prepare and preview fixed announcements — greetings, consent messages, and end-call farewells — as saved audio generated in the background from saved assistant and voice settings, according to the company's changelog and models-and-voices documentation. The prepared audio is reused when a matching combination already exists, but any change to text, voice, language, or relevant speaking settings requires matching audio to be regenerated. Famulor states the preview does not prove playback, acceptance, or legal compliance and does not replace an end-to-end test through the actual phone route, including the mobile network, SIP path, caller device, and consent recognition.", "body_md": "### Summarize Content With:\n\nFixed announcements are small but critical parts of an AI phone call. The greeting shapes the first impression, a consent message needs to arrive clearly and completely, and a deliberate farewell closes the call cleanly. With Famulor’s September 17, 2026 update, teams can prepare these fixed messages and listen to them directly inside the relevant text fields.\n\nThe value is not a blanket speed promise. It is control: before deployment, you can check which voice, language, and wording were actually prepared. This guide covers the complete workflow, the limits around dynamic content, and the real phone-path test that still needs to happen.\n\n**Key Takeaways**\n\n- Eligible fixed greetings, consent announcements, and end-call farewells can be prepared.\n- Audio generation uses saved settings, so save your changes before generating the announcement.\n- A change to the text, voice, language, or relevant speaking settings requires matching audio.\n- A prepared consent announcement does not prove playback, acceptance, or legal compliance.\n- The preview does not replace a test through the actual phone route.\n\n## What are saved announcements in Famulor?\n\n**Saved announcements** are prepared audio variants for fixed moments in a conversation. Famulor generates them in the background from the assistant and voice settings that have been saved. If matching audio already exists for the same combination, it can be reused.\n\nThe current [Famulor models, voices, and saved-announcements documentation](https://docs.famulor.io/assistants/models-and-voices#saved-announcements) identifies three main locations:\n\n1. the greeting on the assistant’s main page and any enabled static message after silence;\n2. the consent message under **Assistant settings → Privacy → Consent** ;\n3. the farewell message in the **end-call tool** editor.\n\nThe [September 17, 2026 changelog](https://docs.famulor.io/changelog#2026-09-17) adds another practical case. If you enter what the assistant should say when a caller declines consent, that consent goodbye can be prepared like the greeting. If the field is empty, the line is generated live. Text chat remains text.\n\nThis makes the feature more specific than a generic TTS preview. You are not listening to a sample sentence from a voice library. You are reviewing the saved wording of a real conversation step with the settings relevant to that assistant.\n\n## What problem do prepared announcements solve?\n\nA voice can sound convincing in a catalog preview and still stress a real greeting incorrectly. Company names, abbreviations, punctuation, long consent wording, and an abrupt farewell often reveal problems only when the actual sentence is spoken.\n\nPrepared announcements move that review earlier. Operations and QA teams can check whether:\n\n- company and product names are pronounced correctly;\n- pace and pauses fit the moment in the call;\n- the consent announcement is understandable and complete;\n- the farewell matches the end-call condition described in the tool;\n- the main language and selected voice fit together;\n- a change correctly triggers the need for matching audio.\n\nThis makes the workflow easier to audit, but it does not make it infallible. The preview represents the prepared audio component. It does not test the mobile network, SIP path, caller device, consent recognition, or the logic that triggers the end-call tool. Those layers still require an end-to-end test. For the broader framework, use the guide to [testing and evaluating an AI voice agent](https://www.famulor.io/blog/how-to-test-and-evaluate-an-ai-voice-agent-in-2026).\n\nIf the wording is not final yet, the existing [AI phone greeting script templates](https://www.famulor.io/blog/11-ai-phone-assistant-greeting-script-templates-make-the-perfect-first-impression) can support the editorial step. This guide begins later in the process: it covers technical preparation and acceptance of wording that has already been selected.\n\n## How do you prepare an announcement step by step?\n\nA reliable workflow is write, save, generate, listen, and place a real test call. Run those steps separately for every critical fixed moment.\n\n### 1. Write a stable, fixed message\n\nUse prepared audio only for content known before the call. A greeting such as “Hello, you’ve reached the digital service assistant at Example Works” is a suitable candidate. A sentence that inserts a caller name or live order status is dynamic and keeps its existing runtime behavior according to the documentation.\n\nKeep the message as short as its purpose allows. For consent, brevity must not remove a purpose that is actually enabled. Famulor’s documentation treats call recording and remembering callers as distinct purposes. The wording required for a specific deployment needs an appropriate subject-matter and, where necessary, legal review.\n\n### 2. Save the text and voice settings\n\nAudio is generated from the **saved** settings. Save the text, main language, voice, and relevant speaking controls first. Otherwise, the preview you expect and the assistant version that is eventually active could diverge.\n\nThis matters when several people edit the same assistant. Record which version has been approved, and avoid parallel changes during the acceptance process.\n\n### 3. Generate audio in the relevant field\n\nThe relevant text fields provide compact controls to **Generate audio**, **Play**, and **Stop**. Preparation runs in the background, so you can continue editing after the save confirmation. The field shows preparation status and, when available, recording duration.\n\nDo more than confirm that the item is marked ready. Listen through the final word. Clipped endings are especially serious in consent and farewell messages because the purpose or the next action may appear at the end of the sentence.\n\n### 4. Review the content and audio separately\n\nA useful review covers at least two perspectives: what the message says and how the saved version sounds.\n\n| Review area | Question | Typical failure | \n|---|---|---|\n| Content | Does the text do exactly what this step requires? | Missing purpose, next step, or contact route | \n| Pronunciation | Are names, numbers, and domain terms understandable? | Wrong stress or clipped ending | \n| Delivery | Do pace, pauses, and tone fit the moment? | Consent too fast or farewell too abrupt | \n| Voice | Does the output match the saved main voice? | Incorrect expectation about the Pipeline fallback | \n| Completion | Does the farewell match the end-call condition? | Polite goodbye without a clear termination path | \n\nChange one relevant factor at a time and generate again. This keeps the cause of an improvement traceable to wording, voice, or a speaking control. For difficult names, numbers, and browser-versus-phone differences, the [practical TTS voice test](https://www.famulor.io/blog/inworld-tts-famulor-voice-agent-guide) provides additional cases without replacing the announcement-specific review described here.\n\n### 5. Test the actual phone call\n\nCall the assistant through the telephony route that will be used in production. Test the greeting, consent, reaction to acceptance and decline, and call ending as one connected experience.\n\nListen for differences between the browser preview and phone audio: softened consonants, changed loudness, clipped word endings, or awkward pauses. Test the logic as well. A perfect farewell recording does not help if the tool description fails to say when the assistant should hang up.\n\n## When is prepared audio reused, and when is it not?\n\nFamulor can reuse matching audio while the underlying combination remains the same. According to the documentation, a change to the message, voice, language, or relevant speaking settings requires a matching recording before prepared audio can be used again.\n\nTreat each of those changes like a small regression test:\n\n1. save the change;\n2. check the new status in the announcement field;\n3. generate audio again where required;\n4. listen to the entire message;\n5. repeat the critical phone path.\n\nThis prevents a risky assumption: “The text was saved, so the old preview is still valid.” An earlier recording may contain different wording or a different voice. The approval record should therefore include the date, voice, main language, and tested phone route—not just the script.\n\n## How do Full Duplex, Pipeline, and dynamic content behave?\n\nThe engine determines which voice is relevant to the announcement. In **Full Duplex**, Famulor uses the native voice of the realtime conversation. The Pipeline voice under **Fallback settings** applies only after the conversation actually switches to Pipeline. The article [Full Duplex in Famulor](https://www.famulor.io/blog/famulor-full-duplex-gpt-live-voice-ai-benchmarks) explains the broader differences between engine modes.\n\nThe current documentation also sets these boundaries:\n\n- Uploaded greetings keep the voice in their recording.\n- Caller-specific variables and other dynamic messages are not frozen into one static audio file.\n- Unsupported modes or unsuccessful preparation keep their existing call behavior.\n- A caller-first greeting still waits for the caller.\n- An eligible fixed message after silence is prepared separately.\n- According to the current changelog, greeting, consent, and farewell use the main language; additional languages are used for switching during the conversation.\n\nCheck the controls available in your current workspace. The interface may vary by engine, voice, and enabled features. This article does not promise universal availability across every plan or configuration.\n\n## Why is a prepared consent message not evidence of compliance?\n\nPreparation and consent are different layers. The audio file shows how a saved message sounds. It does not prove that a specific caller heard it completely, understood it, or accepted it.\n\nThe [Famulor consent and conversation-quality documentation](https://docs.famulor.io/assistants/conversation-quality#consent) describes the runtime check. Depending on the configuration, a caller may respond verbally or with a keypad press. Recording and remembering callers are separate purposes and should not be treated as one. The changelog also confirms that required consent playback completes before acceptance. The separate guide to [cross-channel customer memory](https://www.famulor.io/blog/cross-channel-customer-memory-cx-trend-2026) explains how approved context can be used across channels later.\n\nFor your process, that means:\n\n- Review the wording before generating audio.\n- Test acceptance and decline as separate call paths.\n- Check the call record for the expected result.\n- Do not treat a ready status as consent.\n- Obtain qualified advice for the relevant country, industry, and purpose.\n\nThis article is not legal advice and makes no compliance promise. Technically correct consent handling does not replace an assessment of purpose, legal basis, retention, or access controls.\n\n## Practical example: an [appointment](https://www.famulor.io/use-cases/appointment-booking-faqs)-service call flow\n\nA regional maintenance provider uses a Famulor assistant for appointment confirmations and callback requests. The operations team defines three fixed moments:\n\n- **Greeting:** The assistant identifies the business and the digital service purpose.\n- **Consent:** If recording is intended for this flow, the message names the specific purpose and offers the configured response method.\n- **Farewell:** After a successful confirmation, the assistant repeats only the essential appointment details and says goodbye.\n\nBefore rollout, the team saves the main language and voice, prepares every eligible message, and reviews them with service and privacy owners. It then tests four paths: acceptance, decline, interruption during the announcement, and the normal end of the call. Each path is checked for what the caller hears and what the call record contains.\n\nIf the company name or voice changes later, the previous approval is not automatically carried forward. The team regenerates the affected announcements and reruns the short phone path. This example does not claim shorter calls, higher consent rates, or improved conversion. Those outcomes would need separate measurement.\n\n## How can technical teams manage announcements through API, MCP, or Milian?\n\nThe feature is not limited to the dashboard. Current documentation lists `GET` and `POST` on `/api/v1/assistants/{id}/announcements`. The audio link returned for a ready item is authenticated. MCP and Milian provide `get_assistant_announcements` and `prepare_assistant_announcements`.\n\nThat can support a controlled release check:\n\n1. retrieve announcement status for an assistant;\n2. trigger preparation after an approved text or voice change;\n3. record the result and affected assistant version;\n4. preserve a manual listening and phone test as the release gate.\n\nDo not automate away editorial acceptance. An API can show that audio is ready. It cannot decide whether legally relevant wording is complete for your situation or whether the delivery fits your brand. Famulor says previews are private to the workspace; continue to apply your own access rules to credentials and exported QA records.\n\n## Pre-release checklist\n\nUse a short and repeatable checklist for every production assistant:\n\n- [ ] The main language, voice, and wording are saved.\n- [ ] Only suitable fixed messages were selected for preparation.\n- [ ] Greeting, consent, and farewell were heard through to the end.\n- [ ] Names, numbers, purpose wording, and final words are clear.\n- [ ] Acceptance and decline were tested as separate paths.\n- [ ] The end-call description clearly says when to hang up.\n- [ ] Full-Duplex and Pipeline voices are not confused.\n- [ ] A real call through the intended phone route passed.\n- [ ] Relevant text, language, voice, and setting changes triggered a new review.\n- [ ] Approval, version, date, and responsible owner are documented.\n\n## Conclusion: fixed call moments become reviewable\n\nSaved announcements give teams a clear review point for the most important fixed moments in an AI phone call. You can hear the actual wording with the saved voice, regenerate after relevant changes, and approve the greeting, consent message, and farewell as separate quality components.\n\nThe boundary is just as important. A preview is not a complete call test, prepared consent audio is not acceptance, and a ready status is not legal approval. Combine the audio preview with documented content review and a real phone call.\n\n### Prepare your assistant’s fixed announcements\n\nOpen the greeting, consent settings, and end-call tool in Famulor. Save the intended wording and voice settings, generate the eligible audio, and then test the complete call path.\n\n## FAQ about Famulor saved announcements\n\n### Which messages can I prepare?\n\nCurrent documentation lists fixed greetings, enabled static messages after silence, the consent announcement, and the end-call tool’s farewell. The current changelog also describes a configured goodbye after declined consent.\n\n### Do I need to regenerate audio after editing the text?\n\nYes. A change to the text, voice, language, or relevant speaking settings requires a matching recording before the prepared audio can be used again.\n\n### Do saved announcements work with Full Duplex?\n\nFull Duplex uses its native conversation voice. The Pipeline fallback voice applies only after an actual switch to Pipeline. Test both the preview and the real call in your specific setup.\n\n### Does a prepared consent announcement prove acceptance?\n\nNo. Preparation concerns the audio component. Playback and acceptance are checked in the specific call, and legal requirements depend on the deployment and jurisdiction.\n\n### Can I insert live customer data into fixed audio?\n\nCaller-specific variables and dynamic messages keep their existing runtime behavior according to the documentation. Use prepared audio for genuinely fixed wording.\n\n### Can announcements be managed through the API?\n\nYes. Famulor documents GET and POST for `/api/v1/assistants/{id}/announcements`, plus equivalent MCP and Milian actions. Available permissions and controls depend on your workspace.\n\nWriter at Famulor", "url": "https://wpnews.pro/news/prepare-and-test-famulor-announcements", "canonical_source": "https://www.famulor.io/blog/saved-announcements-famulor-greeting-consent-farewell", "published_at": "2026-09-19 07:00:00+00:00", "updated_at": "2026-09-19 07:53:15.131226+00:00", "lang": "en", "topics": ["ai-products", "ai-tools", "natural-language-processing"], "entities": ["Famulor"], "alternates": {"html": "https://wpnews.pro/news/prepare-and-test-famulor-announcements", "markdown": "https://wpnews.pro/news/prepare-and-test-famulor-announcements.md", "text": "https://wpnews.pro/news/prepare-and-test-famulor-announcements.txt", "jsonld": "https://wpnews.pro/news/prepare-and-test-famulor-announcements.jsonld"}}