Podcastfy vs UNESCO PDF-to-Podcast
Podcastfy
github.comUNESCO PDF-to-Podcast
github.comSide by side
| Figure | Podcastfy | UNESCO PDF-to-Podcast |
|---|---|---|
| Price from | Free | — |
| Free plan | Yes | — |
| Free trial | — | — |
| Open source | — | — |
| Windows | — | Yes |
| macOS | — | Yes |
| Linux | — | Yes |
| API | Yes | — |
| Self-hosted | Yes | Yes |
| Plans published | 1 | 0 |
| Spec score · AI Podcast Generators | 8.0 (#6) | 7.2 (#21) |
| License | The project is licensed under Apache 2.0. | The repository’s LICENSE file identifies the project license as MIT. |
| Requirements | The quickstart lists Python 3.11 or higher and ffmpeg for audio processing as prerequisites. | The README lists Python 3.9 or newer, Poetry, and Ollama as prerequisites, and recommends a GPU while allowing CPU fallback. |
Plans
In their own words
Podcastfy
Podcastfy is a free, open-source Python package that turns multimodal source material into multilingual audio conversations with generative AI. It accepts websites, PDFs, images, YouTube videos, raw text, transcript files and user-provided topics; topic generation can use grounded real-time web search. Users can set the podcast format, style, voices, language and structure, then generate short episodes of 2–5 minutes or longform episodes of 30 minutes or more. Transcript generation supports more than 100 language-model options, including OpenAI, Anthropic and Google models. Listed speech engines include OpenAI, Google, ElevenLabs and Microsoft Edge. Local language models are supported for transcript generation, offering the project’s stated privacy and control benefits. Usage is available through a Python package, command-line interface and beta FastAPI service for URLs, with MP3 audio export. API-key needs depend on selected models; a documented local-model and Edge TTS setup requires none. Python 3.11 or higher and ffmpeg are required. The project uses the Apache 2.0 license.
UNESCO PDF-to-Podcast
UNESCO PDF-to-Podcast turns scientific PDFs into podcast-style conversations. Its pipeline extracts structured material with Docling, uses Ollama with Granite4 to draft a dialogue, then creates audio with VibeVoice. Scripts can target experts, the general public, young people, or novices, and can be generated in English, French, Chinese, Spanish, Arabic, or Russian. Included native voices are available for English and Chinese; for the other four languages, English voices are used unless users provide samples. Users can set the desired duration and choose two to four speakers; defaults are 10 minutes and two speakers. The open-source project is free under the MIT license and can run locally from the command line on Linux, macOS, Windows, or self-hosted setups. Installation requires Python 3.9 or later, Poetry, and Ollama. A GPU is recommended, though CPU use is possible. It currently accepts one PDF at a time, and lengthy documents may need to be split into chunks. Audio generation can demand substantial GPU memory, particularly with the 7B model.