This project is Research Task 06, which builds on earlier work from Task 05 part 1 and part 2.
- In Task 05 - part 1, I analyzed the Premier League 2022–23 dataset without grouping, generated descriptive statistics for the first 10 rows, designed LLM prompts, and compared AI responses with Python results.
- In Task 05 - part 2, I extended the analysis to the full dataset by grouping by stadium, producing pivot-style summaries, targeted visualizations, and Python scripts for more complex queries.
Task 06 takes a creative turn by applying those insights to an AI-generated “street interview” with Arsenal manager Mikel Arteta, reflecting on team performance during the 2022–23 season and looking ahead to the future.
- Create a street interview script based on dataset insights (Task 05 focus: no grouping).
- Generate AI audio interview with distinct voices for interviewer and manager.
- Document multiple approaches attempted, including failures and final workflow.
- Deliver a final merged audio interview simulating a realistic media interaction.
| File Name | Description |
|---|---|
| script.txt | Interview script used as input for audio/video generation. |
| Task_06_Street_Interview.mp3 | Final merged audio interview (from Narakeet + Clideo). |
| Mikel Arteta Discusses Arsenal's Progress and Aspirations-mp4 | Final video interview (Wavel AI) with subtitles and stock footage. |
| experiments/ | Documentation of all attempts (ChatGPT, Gemini, ElevenLabs, Narakeet, Wavel AI). |
| ├── README.md | Detailed log of experiments and outcomes. |
| ├── chatgpt_attempts.txt | Notes from ChatGPT/Gemini trials. |
| ├── elevenlabs_attempts.txt | Notes from voice cloning test. |
| ├── narakeet_clips/ | 10 generated audio clips before merging. |
| └── clideo_merge.txt | Notes on merging Narakeet clips into one audio file. |
- ChatGPT → Could not generate audio/video, but suggested other tools.
- Gemini AI → No audio/video interview support.
- ElevenLabs → Successful voice cloning of Arteta using sample from
voicy.network, but full TTS required paid subscription. - Narakeet → Generated 10 separate clips (Interviewer + Arteta).
- Clideo.com → Merged clips into single interview.
- Python (pydub) → Attempted merging locally, but required FFmpeg installation. Not pursued.
- WavelAI → Generated a video Interview.
- Drafted street interview script (Interviewer + Mikel Arteta).
- Tested multiple platforms (ChatGPT, Gemini, ElevenLabs, Narakeet).
- Produced audio-only interview with Narakeet + Clideo.
- Produced a video interview with Wavel AI, including auto-subtitles and contextual football clips.
This task demonstrated:
- How analytical insights can be transformed into narrative storytelling.
- The process of evaluating and troubleshooting multiple AI tools (balancing free vs. paid features).
- How different platforms add different value:
- Narakeet gave controlled dialogue (audio only).
- Wavel AI added creativity, recent football references, and stock visuals for a realistic video.
The final result is a combination of audio and video deliverables, showing both technical adaptability and creative application of AI tools.