PDF to Podcast

Overview

This Solution Blueprint provides an end-to-end pipeline for converting PDF documents into podcast-style audio content. It uses agentic orchestration with microservices for PDF processing, LLM-powered script generation, and Text-to-Speech synthesis.

AMD Solution Blueprints are packaged as Helm charts for deployment on a Kubernetes cluster. For development or further exploration, the source code is public and available in the Solution Blueprints GitHub repository.

Architecture

The blueprint integrates a Gradio frontend, a FastAPI application with Celery workers, an AIM LLM service, a Qwen TTS service, and Redis for task queuing. By default, the AIM deploys Llama 3.3 70B for script generation and a Qwen TTS model for audio synthesis.

Component	Role
Frontend (Gradio)	Web UI for uploading PDFs and managing conversions
App (FastAPI)	REST API that coordinates the pipeline
Celery worker	Background processing for PDF extraction, LLM calls, and TTS
AIM LLM	Script generation (default: Llama 3.3 70B)
Qwen TTS	Text-to-speech synthesis
Redis	Task queue and message broker

Key Features

• End-to-end pipeline: Ingest PDFs, extract text, summarize/plan, generate dialogue or monologue with LLM, synthesize audio via TTS, and store artifacts

• Agentic orchestration: Agent service coordinates PDF/TTS/LLM calls, uses Redis for tasks and MinIO for artifacts

• Multiple modes: Podcast dialogue or monologue, controlled via request parameters

• Flexible LLM and TTS configuration: Deploy bundled AIM services or connect to existing endpoints

Getting Started

This is a quick start guide on how to deploy the blueprint. For advanced options, such as reusing an existing AIM, providing a Hugging Face token, or overriding storage classes, see Deploying Solution Blueprints with Helm or explore the advanced deployment guide.

This blueprint supports AMD Instinct (default) and AMD Radeon platforms. The section below covers the default Instinct deployment. For Radeon and other advanced options, see:

• Deploy on AMD Radeon

Prerequisites

System Requirements

The blueprint requires the following cluster resources by default:

Resource	Default Configuration
GPUs	2
CPUs	16 CPU cores
RAM	104 Gi

Model serving GPU requirements (default in this Blueprint):

• LLM model (amdenterpriseai/aim-meta-llama-llama-3-3-70b-instruct via aimchart-llm): 1 GPU (amd.com/gpu: 1)

• Speech model (Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice via aimchart-qwen-tts): 1 GPU (amd.com/gpu: 1)

• Total when both internal services are enabled: 2 GPUs (1 for LLM + 1 for TTS)

• If you use llm.existingService and/or qwen-tts.existingService, GPU requirements for those external services are defined by their own deployments.

Persistent storage defaults (from values.yaml in the repository):

• Ephemeral storage: 20 Gi (ReadWriteOnce) for temporary workloads

• App storage: 10 Gi (ReadWriteMany) for shared PDF and temp data

• Shared memory (dshm): 32 Gi for /dev/shm

Note: Celery-worker resources should be increased to handle OOM issues during heavy PDF processing and LLM tasks. The specified memory requests for celery-worker are suitable for processing PDF files up to 20 MB in size. For larger files, significantly more memory may be required (proportionally to file size).

To deploy to the Kubernetes cluster, ensure the following prerequisites are met:

• kubectl: Installed and configured to communicate with the cluster

• Helm 3.17 or higher: Installed on your local machine

Deployment

Solution Blueprints are packaged as OCI-compliant Helm charts in the Docker Hub registry and can be deployed to a Kubernetes cluster with a single command. Define the name (deployment name) and the namespace (Kubernetes namespace), then pipe the output of helm template to kubectl apply -f -:

name="my-deployment"
namespace="my-namespace"
helm template $name oci://registry-1.docker.io/amdenterpriseai/aimsb-pdf-to-podcast \
  | kubectl apply -f - -n $namespace

Note: You can create a namespace using kubectl create namespace $namespace.

To check the status of the deployment, run:

kubectl get pods -n $namespace

Wait until all pods report Running and Ready. The default AIM LLM is large and can take several minutes to start.

Connect to UI

To connect to the UI, port-forward to 7860. The UI will then be available at http://localhost:7860 in your browser.

kubectl port-forward services/aimsb-pdf-to-podcast-${name}-frontend 7860:7860 -n $namespace

Once connected, use the application as follows:

1. Upload a target file: The main PDF document that will be converted into a podcast. Exactly one target file is required; the pipeline extracts text from it, generates dialogue or monologue, and optionally synthesizes audio.

2. Optionally upload context files: Optional additional PDFs that provide extra information to the LLM. They are used as reference when generating the podcast from the target document (e.g. terminology, or background). You can upload multiple context files.

3. Choose conversion options:

   • Monologue Only: Single speaker, no dialogue

   • No TTS: Skip audio generation; output is transcript only

   • Full audio: Generate the full podcast; otherwise audio is limited to 3000 characters

4. Start the conversion and download the transcript and audio when complete.

If No TTS is selected, audio will not be generated even when Full audio is enabled.

Clean Up

When you are finished, remove the deployed resources:

helm template $name oci://registry-1.docker.io/amdenterpriseai/aimsb-pdf-to-podcast \
  | kubectl delete -f - -n $namespace

LLM model compatibility

This blueprint is validated with LLM backends exposed via llm.existingService or the default aimchart-llm dependency. The application is designed to operate correctly with models of capability level not lower than Llama 3.3 70B. Prompts and pipeline configuration have been tuned and tested for this class of models. Using smaller or less capable models may lead to issues such as incorrect or unstable structured output; in such cases, additional prompt or configuration tuning may be necessary.

Third-Party Components

This Solution Blueprint uses multiple third-party components. To see the full set of software and Python dependencies, explore the repository source and dependency files. The table below highlights some of the key components. For further license information, refer to each component’s official documentation.

Component	License
Celery	BSD
FastAPI	MIT
Gradio	Apache 2.0
LangChain	MIT

Terms of Use

AMD Solution Blueprints are released under the MIT License, which governs the parts of the software and materials created by AMD. Third-party Software and Materials used within the Solution Blueprints are governed by their respective licenses.