Practice speaking about a photo, then review corrections
At a glance
A language learner configured a photo-questioning GPT and used voice mode for structured Russian practice. This workflow is listed under Language Learning, Education.
- Agents / tools
- ChatGPT, Custom GPTs, Voice mode
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Use an account supporting custom GPT creation, image input and mobile voice chat. In the GPT's configuration, instruct it to ask for the target language, proficiency and a photo, waiting for each answer. 2) Have it ask the learner to describe the photo in that language, correct errors briefly, and ask two further questions with feedback after each answer. 3) Start in text to supply the language, level and image, then switch to voice and answer aloud. 4) End with the GPT's model description and a complete English translation of the exchange. If translation stops early, ask it to continue.
How it went
The author's third image-based trial worked better using beginner-level Russian. Asking it to continue resumed a prematurely stopped translation.
Worth knowing
This is a February 2024 experiment. Trials included image-recognition failure and mistaken language detection; current interfaces and access can differ.
