ELEVEN REALTIME TRANSLATION: TWO-DIRECTION WORKSHEET
Original free planning resource | Prepared October 25, 2026

No provider connection, audio upload, translation, recording or paid call was
performed for this resource. All examples are fictional and untested. Reading
the resource is free; enterprise access and production services are separate.

1. CHOOSE THE HUMAN CONVERSATION
Business task:
Speaker V / role:
Speaker S / role:
Languages and qualified reference reviewers:
Who can explain the relay and handle an unavailable segment?
Allowed use and recording/retention decision:
What decisions remain with the human participants?

Do not turn a translated request into a reservation, instruction acceptance,
payment, staff notification or other action without its own verified result.

2. ACCESS AND COST READINESS
Enterprise workspace / responsible owner (no secrets):
Alpha access confirmed from:
Required account terms and funding reviewed by:
Enterprise price and billable-unit evidence:
How two directions/sessions are charged:
Other network/application/carrier/storage costs:
Allowed future test scope / spending owner:
Status: DESIGN ONLY / ACCESS UNKNOWN / READY FOR AUTHORIZED REVIEW

No public numeric translation tariff is supplied here. Do not reuse a price
for a different Realtime voice-agent model or speech-generation product.

3. ORIGINAL TWO-DIRECTION CONFIGURATION INVENTORY
Example purpose: a human visitor asks exhibition staff about public entry.
Suggested language pair for qualified review: Spanish and English.
This is a design choice, not evidence of accepted or accurate service.

Direction A: V -> S
Source: visitor microphone; source language Spanish, if confirmed
Target: staff member's reviewed English
Source encoding, actual evidence:
Selected target language / output voice:
Chosen TTS model and account eligibility:
Eagerness setting:
Playback destination: staff listening channel only
Application session / provider session mapping:
Raw-input versus translated-output separation:
Status: NOT CONNECTED

Direction B: S -> V
Source: staff microphone; source language English, if confirmed
Target: visitor's reviewed Spanish
Source encoding, actual evidence:
Selected target language / output voice:
Chosen TTS model and account eligibility:
Eagerness setting:
Playback destination: visitor listening channel only
Application session / provider session mapping:
Raw-input versus translated-output separation:
Status: NOT CONNECTED

Keep the provider credentials in an approved protected component, never here.
Confirm current endpoint fields with the API reference rather than copying
an unreviewed URL or an example voice ID into production.

4. ORIGINAL MEANING FIXTURE - NOT A TRANSLATION RESULT
Visitor:
I meant the Saturday exhibition, not the Friday workshop. Can I arrive through
the east entrance after eleven?

Staff:
The exhibition is Saturday, November 14, 2026. Public entry starts at eleven
thirty through the east entrance. The Friday workshop has separate entry
instructions.

Qualified reviewers must prepare and check the intended languages. The English
script above is planning material, not an independently verified translation.

Required meaning in the final exchange:
- Exhibition is Saturday, November 14, 2026; Friday workshop is a different event.
- Public entry is 11:30, not 11:00, through the east entrance.
- Arrival question does not become a booking or guaranteed admission.
- The correction remains attached to the visitor's request.

Reference version / source recording, if later approved:
Who checked actual spoken words and the translation meaning?

5. RETAIN THE EVIDENCE FOR EACH DIRECTION SEPARATELY
Direction / attempt / session reference:
Actual source audio reference, if authorized:
Interim source text and superseded details:
Final source text:
Translated text:
Audio-output evidence:
Recipient observation, including silence or interruption:
Meaning reviewer / reference version:
Expected day/time/entrance recovered correctly? Evidence:
Failure or uncertainty kept visible:
Decision: ACCEPT FOR THIS CASE / REVISE / UNRESOLVED

A text event is not proof of audible delivery. A continuing session is not
proof that a failed segment was translated. Preserve failed attempts.

6. ORIGINAL FAILURE CASES - ALL UNPERFORMED
T1 - Visitor changes Friday to Saturday
Expected: Correction survives; no answer based on the superseded Friday event.
Observation: NOT RUN

T2 - Entry time distinction
Expected: Recipient recovers 11:30; eleven and eleven thirty are not accepted
as equivalent for this fixture.
Observation: NOT RUN

T3 - Caption exists, audio missing
Expected: Operator sees unavailable output; never labels text alone as heard.
Observation: NOT RUN

T4 - One segment fails while connection remains
Expected: Mark the missing content; no stale successful answer substituted.
Observation: NOT RUN

T5 - Session closes
Expected: Stop claiming an active relay; use an approved human next step.
Observation: NOT RUN

T6 - Playback enters the source microphone
Expected: Fix application routing; do not treat echoed translation as a new
participant statement or a successful second direction.
Observation: NOT RUN

7. CONTROLLED EAGERNESS REVIEW PLAN
Reference / audio / voices / directions held constant:
Current setting:
Single proposed change:
Meaning failures observed:
Delay or conversation-flow failures observed:
Attempts retained, including failures:
Decision owner / next action:

Do not choose a universal setting from one greeting. This worksheet contains
no measured quality score, numeric latency promise or ranked setting.

8. ACCEPTED SCOPE AND BURKI BOUNDARY
What was configured:
What was actually accepted by provider:
What each recipient actually received:
What remains unknown:
Two-way route, language pair and device scope accepted by:
Follow-up owner / review trigger:

This is a direct-provider implementation plan. A native Burki connector to
Eleven Realtime Translation is not established by this guide. Existing speech
provider components do not establish this service's availability.

Companion guide:
https://burki.dev/blog/eleven-realtime-translation-two-way-setup
Primary API reference:
https://elevenlabs.io/docs/api-reference/realtime-translation/v-1-translation-realtime