AI · Generative Media

Trained on Adobe Stock and public-domain work, so commercial legal risk is low. It gets picked for client work because of that licence clarity, not image quality.

Levels, denoises and normalises loudness to broadcast targets (LUFS). A post-production tool, not a live one — the last step before a podcast ships.

A list of tools and resources for Whisper, the open-source speech-recognition model. Unlike cloud speech services you can run the model locally; the list gathers accelerated ports and front-ends.

DALL·E-based with no barrier beyond a Microsoft account. Not top-tier quality, but among the lowest-friction options.

Translates a natural-language instruction into Blender Python API calls. It does not generate models — it operates Blender’s own tools for you, so the result stays reproducible.

A template library and brand kit produce consistent output without design skill. Unlike Figma it is template-first; that is also the ceiling — the tool resists once you step outside the templates.

Auto-captioning, speaker tracking and vertical-format templates. Far faster than Premiere for short-form, and correspondingly shallow on colour and audio.

Bundles object removal, relighting, upscaling and background replacement in one place, so one account covers what would otherwise be several services.

You build the pipeline as a node graph — model loading, sampling and VAE decode are separate nodes. Unlike form-based interfaces such as AUTOMATIC1111, a workflow exports as JSON and reproduces exactly; the price is a harder setup.

An image generator that works without signup. Compared with current models its output quality is well behind; it claims nothing beyond fast, frictionless experimentation.

Generates textures and HDRIs inside Blender, with seamless tiling and projection painting straight onto a mesh. Instead of generating elsewhere and importing, everything stays inside Blender.

The opposite pole from ComfyUI — it hides sampler, steps and CFG behind good defaults, aiming to deliver a Midjourney-like experience locally.

Produces scrollable pages rather than slides, with the layout adapting as content grows so you stop fighting to fit a box. Exports to PowerPoint and PDF.

An interior design tool. Unlike free-form 3D modelling tools it works from catalogues of real furniture brands, so every item in the scene has a purchasable counterpart.

One of the few models that renders legible text inside an image. Unlike other diffusion models, it gets letters right on posters, logos and packaging where they mangle them.

Generates intermediate views from one image via multi-view diffusion, then reconstructs a mesh. Returning usable surfaces rather than a point cloud shortens the rest of the pipeline.

Browser-based video creation platform that generates scripts, voiceovers, stock footage scenes, and cuts from text prompts; modifies edits via natural-language commands unlike traditional template editors.

Regional generation and editing on an infinite canvas, built for professional workflows and team use. Unlike ComfyUI’s node graph it is canvas-first, so you work like a painter; a commercial edition exists too.

Installs as a virtual microphone device and kills noise at the source. Unlike per-app noise suppression, it works across Zoom, Teams and any recorder at once.

Wires Krita’s layers and masks straight into diffusion — select a region and generate into it. You stay inside your existing painting flow instead of switching tools.

An image-generation platform. Unlike Midjourney you can train a style model on your own images and reuse it — decisive when game assets need a consistent look.

Alongside video generation it turns an object you filmed on a phone into 3D via NeRF/Gaussian splatting. Unlike pure video models, it has that second leg.

Generates a 3D model from text or an image and applies PBR textures itself. Unlike tools that produce geometry alone it textures too — the slowest step in an asset pipeline.

Returns the generated image as a layered, editable layout, so unlike the flat raster you get from pure generators you can keep working on it.

Turns plain prose into flow charts, timelines and comparison graphics as editable vectors. Unlike deck generators, it visualises one idea rather than building a whole deck.

Joins a meeting, transcribes live and separates speakers. Unlike upload-after-the-fact tools it lets you stop taking notes while the meeting is still running.

An image generator focused on character and product imagery. Unlike general generators it has that focus, plus a free tier generous enough for daily use.

Cloud video generator that turns articles, scripts, and webinar recordings into captioned short-form videos; edits by matching text segments to stock footage rather than manual timeline cutting.

Generates short video from text and images, with camera motion and effects exposed as parameters — you steer the motion instead of hoping the prompt lands.

A presentation tool with live co-editing, comments and brand templates, so a deck behaves like a team document. Unlike PowerPoint it has analytics, showing which slide a recipient stopped on.

Draw a floor plan and get an instant 3D render. No CAD precision — it is for a fast presentation image, not a dimensioned technical drawing.

Combines generation and editing on one canvas, with region-selective regeneration (inpainting) built into the flow. Unlike generate-only tools, it suits piecemeal correction.

OpenAI’s 3D model that produces point clouds needing a separate meshing step. Compared with its successor Shap-E it is fast but coarse; little practical use is left today beyond historical interest.

Generates visuals and short video alongside copy, with competitor-oriented templates. That is where it separates from text-only tools.

Can output SVG. For logos, icons and illustration that is the fundamental split from raster generators — the result scales and stays editable.

Single-purpose background removal with an API and batch mode, so it drops into a pipeline. Better than general tools on hair and transparent edges.

An OpenAI model producing implicit 3D representations from text. A research release: the output is not production quality, and its value is historical and academic.

Bundles design, copy, video editing and scheduling in one app. Compared with the specialists it trails at each task, but it is a sensible trade if reducing tool count is the goal.

A Google Slides add-on that generates slides from text. Instead of moving to a separate presentation platform you stay in your existing file, with no export step.

Design a 3D scene in the browser and embed it straight into a page. Not general-purpose like Blender — the output is optimised for the web from the start.

The most widespread interface for local image generation, with a huge extension ecosystem (ControlNet, LoRA loaders). Unlike ComfyUI it is form-based rather than a node graph; the repo is no longer actively developed, but its installed base is still the largest.

Produces a UV-unwrapped, textured mesh from one image in about half a second. Quality is not top-tier, but at batch scale the speed difference decides.

Turns text or a rough sketch into UI and hands back an editable design file — unlike the code generators, output a designer can take over.

A text- and image-to-3D tool, a direct rival to Meshy. Compared with Meshy, generation is faster and topology generally cleaner, which means less manual cleanup afterwards.

A framework that upscales video frame by frame with neural models, driving waifu2x, Real-ESRGAN and Anime4K behind the scenes. Against single-image upscalers it takes on the video pipeline and temporal consistency.

Bundles voice cloning, dictation and generation into one open-source studio. Keeping the voice sample off a cloud service is a real privacy difference for cloning.

Open-source voice synthesis studio combining text-to-speech, voice cloning, and multilingual dubbing in one interface; runs offline on a local GPU unlike paid cloud voice APIs.

The most widely used open speech-recognition model, covering 90+ languages at usable accuracy in Turkish too. Unlike cloud speech services it runs locally, so the recording never leaves the machine.