Task: Program for automatic video assembly (local, for personal use)

Budget: 27000 UAH Deadline: 15 days

Sergiy, the task is complex but very well formalized — this is exactly the kind of pipeline I would take on. I will create a local Windows program in Python with a modular architecture: LLM script, ElevenLabs, connection to 6+ API sources, fallback search, frame verification through LLM, assembly via ffmpeg/moviepy, auto cache cleaning, and disk limits. I have experience with API integrations, media processing, and GUI. For such a project, local files/JSON will be sufficient; a database is not necessary. I can immediately propose the GUI structure and implementation stages.

Cost: 20,000
Deadline: 15

Budget: 27000 UAH Deadline: 14 days

Sergey, good afternoon. I will respond immediately to your four points, then about the main risk of the technical specification.

Work examples. I do not have a public link specifically to the ffmpeg video assembly - I have done similar work as internal scripts for clients, I cannot show them, and I will not present fabricated cases. From the open ones - accessibilityscanner.top, my live service for automatic website audits: the same mechanics you need - task pipeline, API integrations, a ready report at the output. I constantly work with OpenAI/Anthropic API, have dealt with stock APIs (Pexels, Pixabay) and ElevenLabs - the latter has an endpoint with-timestamps, I will explain below why you need it.

GUI. Tkinter with standard ttk: zero external dependencies, double-click works without installing Qt. The window is simple - theme, list of voices (pulled from ElevenLabs at startup), option checkboxes, progress with a log, Create button. If you want PyQt - I will do it, but for one user, it’s unnecessary weight.

Database. SQLite, one file next to the program, nothing needs to be installed. You cannot do without clean files: a database of used clips is needed for uniqueness and an index cache with the date of the last access - without it, auto-cleaning by limit in gigabytes turns into a lottery, you can delete a clip that is currently being assembled.

Deadline and cost. 14 days, 27,000 UAH, within budget. I will provide a working version with 3-4 sources for testing in about 8-9 days, the remaining sources and polishing in a second round - you will have time to test it on your topics.

Now the risk. Timing. The length of the scene should be dictated by the audio, not the other way around: first the voiceover, we take the exact duration of each piece and cut the clips accordingly and calculate Ken Burns. If you assemble the video sequence and adjust the voice - by the tenth scene, desynchronization by seconds is guaranteed. Plus, I will normalize clips from nine different sources to one fps and pixel format before stitching, otherwise ffmpeg will struggle with the concat. And subtitles from here as well: with-timestamps provides character-level timings, .srt comes out accurate without speech recognition.

If the approach is okay - write, I will look at your references and clarify details about the intro teaser. Expand

Budget: 27000 UAH Deadline: 10 days

Hello. The architecture of the faceless video generator is transparent: a modular pipeline in Python from the LLM script to rendering through ElevenLabs and stock API.

Interface and storage: PyQt6 for GUI. A heavy database is not needed; for cache control and keeping a history of used media, a local SQLite is sufficient.

Rendering: Direct calls to ffmpeg (works more reliably and faster than moviepy for the Ken Burns effect and dynamic stitching of different format scenes).

Background: I have implemented complex asynchronous systems with integration of multiple APIs (trading terminals, HFT bots). The fallback logic for switching between 6+ sources and validating frames through LLM will be fault-tolerant, and the modules will be easily replaceable.

What is the average duration planned for the final videos to calculate the load on the system during local assembly?

Budget: 27000 UAH Deadline: 14 days

Serhiy, you need a local tool for automating the video creation pipeline so you don't waste time on manual editing of typical clips. I will set up a script that combines the OpenAI API for scripts, ElevenLabs for voice, and searches through Pexels or Pixabay for automatic visual selection using MoviePy or similar libraries for stitching. The program will work with a simple interface on your PC, where you can insert keys and manage the generation. What volume of video do you usually plan to create at one time to correctly set the generation timings and API limits?

Budget: 27000 UAH Deadline: 10 days

Hello! The task is clear, it involves a Python script with integration of OpenAI/Anthropic, ElevenLabs, and free stock API. I will implement script generation, visual selection, voiceover, and final assembly in MP4 according to your rules.

Message me, we will discuss everything in detail, and I can start working immediately.

Portfolio: Freelancehunt
Reviews: Freelancehunt

Budget: 27000 UAH Deadline: 15 days

Hello, Sergey.

I have read the technical assignment in full. The project is realistic, but I would build it not as one large Python script, but as a modular local application with separate adapters for LLM, ElevenLabs, each media source, and the video assembly engine.

For the interface, I suggest PySide6. This will allow for a proper Windows application with settings, voice selection, progress by scenes, execution log, information about occupied space, stop, restart, and cache clearing buttons.

For state storage, I propose using a local SQLite database. It will store:

— created projects and the processing state of each scene;
— IDs and hashes of used photos and clips;
— source, author, URL, and license of each material;
— information about temporary files and cache;
— history of errors and retries.

An external server is not needed for this. All data and finished videos remain only on the user's computer.

I suggest performing video assembly through FFmpeg. MoviePy can be used for individual auxiliary operations, but the main processing, format normalization, scaling, transitions, audio mixing, and final export are more reliably done directly through FFmpeg.

The architecture will be divided into modules:

— LLM providers: OpenAI and Anthropic;
— TTS provider ElevenLabs with automatic retrieval of the list of available voices;
— a separate module for each media source;
— analysis and breakdown of the script into scenes;
— selection and verification of visuals;
— video sequence planner;
— generation of intro teaser;
— subtitles and background music;
— video rendering;
— cleaning up temporary files and cache control.

I propose the following order of implementation.

Stage 1. Working core and prototype:
— GUI;
— settings and API keys;
— generation or import of the script;
— breakdown into scenes;
— obtaining voices and voiceover through ElevenLabs;
— integration of the first two media sources;
— assembly of the first test MP4.

Stage 2. Media sources and licenses:
— connecting the remaining sources;
— fallback between them;
— filtering by license type;
— saving author, license, and URL;
— randomization of results;
— database of previously used materials.

Stage 3. Complete video pipeline:
— cutting clips to the duration of the voiceover;
— Ken Burns for photos;
— transitions;
— control of frame duration;
— planning the share of video and photos;
— intro teaser from the most important scenes;
— subtitles and music;
— final export 1920×1080.

Stage 4. Stabilization and transfer:
— automatic cleanup of temporary files; Expand

Budget: 27000 UAH Deadline: 20 days

Good day!

I have relevant experience: I built microservices with LLM API integration (FastAPI + OpenAI, report generation, working with API keys through configuration) — the same logic of modular integration of external services is needed here (ElevenLabs, free media sources, fallback between them).

GUI: I propose PyQt — more flexible than Tkinter for the progress bar during generation (script → voiceover → video selection → assembly — each stage is long, status needs to be shown), plus it's easier to make a neat window appearance.

DB: local SQLite will be sufficient — for the database of already used clips (uniqueness) and tracking cache (size, upload date for LRU cleaning). An external DB is redundant here for a single-user desktop application.

I am ready to discuss the price range and timelines after clarifying the scope (how many sources to implement at once in the MVP, whether subtitles/music are needed in the first version, or if they can be added later).

Budget: 27000 UAH Deadline: 7 days

Automatic cleaning of intermediate files after assembly is a critical detail that is often overlooked in similar projects, but it prevents disk clutter.

I will do the following: a Python script with a modular architecture, where each source (Pexels, Pixabay, Wikimedia, Archive.org, and others) is a separate module with a unified interface; LLM (OpenAI/Anthropic) generates the script and breaks it down into scenes with search queries; ElevenLabs API voices the text with a selected voice; for each scene, the program sequentially searches for video/photos, checks through LLM (max. 3 attempts), then ffmpeg/moviepy stitches everything together according to the timing of the voiceover with a Ken Burns effect for photos; GUI on PyQt5 for entering the topic and selecting the voice; all API keys in a config file; after the finished MP4, all temporary files are automatically deleted, and the database of used clips prevents repetitions.

I offer 2 weeks of free code support after delivery — the program is complex, and debugging may be needed when working with different APIs and handling non-standard topics.

Ready to start.

Budget: 27000 UAH Deadline: 3 days

Hello!

You have a wonderfully written, rare in quality technical specification. The logic of the application's operation is completely clear. I specialize in process automation using Python, working with neural networks (OpenAI / Anthropic), and software processing of media files (FFmpeg / MoviePy).

Budget: 27000 UAH Deadline: 30 days

Good day
I can develop this for you
Write to me

Budget: 27000 UAH Deadline: 10 days

💫 Hello! Sergey Dumенко.
I have created similar automation pipelines in Python, combining LLM script generation, ElevenLabs voice synthesis, and video compilation using MoviePy/FFmpeg with Ken Burns effects and transitions.
For the graphical interface, I propose Tkinter as a simple window that launches with a double-click, featuring a list of voices obtained from the ElevenLabs API.
I will use a local SQLite database to store used media hashes for tracking uniqueness and managing the cache, with automatic cleanup of temporary files after each rendering and a customizable cache limit with a manual clear button.
The modular architecture will separate each media source (Pexels, Pixabay, Wikimedia Commons, etc.) as interchangeable plugins, with LLM backup logic limited to 3 checks per scene to control costs.
I estimate the total development time to be 2 weeks and the cost to be $1000, with fully commented source code and a README file for clean setup from scratch.
All rights will transfer to you after final payment.

Budget: 27000 UAH Deadline: 5 days

Hello!
I am ready to complete your project. I have reviewed the description, and the task is clear. I have experience in developing similar solutions, quickly understand the requirements, and deliver results.

We can discuss the details and get started today.

Budget: 27000 UAH Deadline: 1 day

Hello. I will develop a program with a clear modular architecture in Python, using PyQt for an intuitive interface that will allow easy integration of new sources and LLM providers. The main focus will be on integrating LLM for generating scripts and verifying visual content from various APIs, and for caching and tracking already used media files, I will use a lightweight SQLite database. I have significant experience with ffmpeg/moviepy for efficient video assembly and optimizing disk space management, ensuring high performance and resource conservation, and I already have ready-made solutions. The project will be implemented with an emphasis on code cleanliness, detailed documentation, and independence from the developer, ensuring easy maintenance and future modifications. I suggest discussing all implementation details, final budget, and deadlines in private messages.

Budget: 27000 UAH Deadline: 21 days

Hello! I can develop a program for automatic video assembly in Python, using the necessary APIs and libraries. I have experience working with ffmpeg, ElevenLabs, and LLM. I am ready to discuss the details and propose an optimal solution.