Create, edit, mix and design in one window, with local models and full MCP support.
All the local models you need, with no setup required.

Every picture, clip and sound on this site was made in TinyPine












































Everything on this site is in the free app: no paid tier, no locked features, no trial period, no subscription and no account. For personal and commercial work, and version 1 stays free for good.
Pick what you came to do. Each area has its own page with what it can do, shown in the app. They share the library, the queue, the undo and the grade, so work moves from one to the next without exporting anything.









Every line is a path the app really has. Pick an area to see what it hands on and what it takes in.
Prompt, Assist, Vary takes and your own LoRAs.
A local language model with 26 styles and a dial from Faithful to Wild. Each example is a real run: the line, what Assist wrote, and what Krea 2 made of it.
“an old fisherman mending his nets at dawn”
Assist · Portraitan old fisherman mending his nets at dawn. Hands knotted with salt and time, threadbare nets draped low, taut and frayed across his knees. Low-angle close-up, frame tight on labor, off-center to let the tide’s edge breathe. Dawn’s hush over tidal shore, mist curling at water’s edge, light soft and golden from the east, catching grit and cloth. Muted blues, ochres, steel grey, worn by wind and sea. Photographic portrait, natural light, fine grain, shallow depth of field
waiting for the prompt
Attach a picture or a video frame and ask what to do next. Assist describes what it sees, proposes the next scene and writes its prompt. The answer here is the one Assist 8B gave about a dancer in a pine forest; the Images page has it doing the same for two fox cubs, and the render that came of it.
TinyPine Studio has MCP servers for the timeline and for Design. Connect Claude Code, Claude Desktop or any client that speaks MCP, and the agent can do almost everything you can do in the app: make pictures, video, music, voices and sound effects in the app's own queue, edit and grade them, cut a film and export it finished, lay out a page, run the library and train a LoRA. An open timeline or design shows each change within two seconds as an undo step, so Ctrl+Z takes it back, and what you change at the same time is merged with what the agent does. Moving the library stays with you.
121 tools in the timeline server and 19 in the Design server. These are the ones an agent reaches for most, by their real names.
Render with LoRAs and Enrich, vary a picture, upscale it, fix faces, cut out the subject and edit it with Auto, presets or any Edit slider.
make_imagevary_imageenrich_promptvary_promptassist_chatedit_imageupscale_imageenhance_facesdepth_mapcut_outselect_objectlens_profilephoto_mergeexport_with_editsA clip from a picture, longer with Extend, slower, steadier, twice the size, or with something tracked in it.
make_videoextend_videoslow_motionstabilize_videoupscale_videotrack_motionSongs and longer songs, voices and saved voices, sound effects, and the words in any clip.
make_musicextend_musicmake_speechsave_voicemake_sound_effecttranscribeadd_subtitlescheck_scriptcheck_singingTimelines, tracks, clips, titles, panels and shapes, placed where the agent says.
create_timelineadd_trackadd_mediaadd_textadd_graphicadd_shadowadd_layersupdate_clipdelete_clipimport_mediaKeyframes, text letter by letter, transitions, masks, green screen and effect clips.
set_keyframesset_animatorsapply_text_presetset_transitionset_masksset_chromaadd_effectSee any frame, check the cut, export the finished video, or hand it to Final Cut Pro, Resolve or Premiere.
render_framecheck_before_exportexport_timelineexport_interchangeexport_filesSearch by what a picture shows, rate, label, sort into bins, follow where a file came from, and restore from the trash.
search_libraryitem_inforateset_titlelabel_itemsmake_binbin_itemslineagewrite_xmpdelete_itemrestore_itemCaptions, training, a sample per checkpoint and the one to keep.
train_loradatasetauto_captionset_captionlora_samplesprefer_checkpointlist_lorasWatch and steer the queue, install a model from the Grove, read and change settings.
list_queuecancel_jobretry_jobpause_queuelist_modelsinstall_modelget_settingsset_settingsThe order of work, which model does what, and every option by name.
video_workflowmodel_guidemusic_guidelist_optionslist_presetslist_edit_settingsArtboards, text, shapes, pictures, effects, groups, masks and brush strokes, with the artboard exported at full size.
create_designadd_artboardadd_textadd_shapeadd_imageadd_fxset_fxgroup_layersset_maskpaintupdate_layerdelete_layerset_paletterender_previewexport_artboardWindow › Connect an AI agent in the timeline and in Design shows the exact command for Claude Code and the settings for Claude Desktop, with Copy buttons. The servers run on your computer and read only your library.
TinyPine Studio is a native Windows app written in Rust on GPUI, the GPU-accelerated interface framework made for the Zed editor. Every frame of the interface is drawn with Direct3D 11, and the effects run on the same card. Panels fold for the picture, rows fold step by step instead of wrapping, and dialogs never leave the window. Before every build the app opens each view at four window sizes and four zoom levels and checks that nothing spills out.

1600 × 1000 app pixels · zoom 100 % · no element outside its box, checked by the app itself
Auto reads the light, the key and the colour of a picture and grades it toward the look you pick: Standard, Cinematic, Dark & moody, Vibrant or Soft & airy. It says why it changed each thing, and you can take back any part. Pick a picture and a look, then drag across it.
Cinematic. Film-like: soft lifted blacks, highlights rolled off, a little darker, quieter colour, a warm lean.

OriginalAuto · CinematicA complete audio workstation inside the video editor. A channel strip per track with EQ, compressor, Auto tone, Tame and a limiter, send buses with reverb and delay, and a master that measures loudness against the standard you deliver to. The mix you hear is the mix the export writes. This is the film from the Editing page, playing.


Shape the tone with up to eight bands on a live spectrum. Drag a point to move a band, double-click the graph to add one.
The same evening, the same library: a line becomes a picture, a grade, a clip with sound, a song, a cut and a poster. Every step below is the real app.








Write it the way you would say it, pick one of 26 styles and how freely Assist may read it.




















One recipe, one set of shaders. Grade a still and the clip made from it gets the same look; on the timeline every setting can be keyframed, and in Design an FX layer applies it to the page. On video the light effects read every frame: flares follow the bright sources and grain is new on every frame.

Vary asks Assist for a new take of the same prompt: another pose, place, light or weather. Who is in the picture and how it is made stay locked, and every take is checked against the original before it renders. Queue as many rounds as you like and let it run over the weekend.









Each clip below started as one of the pictures on this site. LTX-2.5 animated it and wrote the sound in the same pass. Click a clip to hear it.
Extend continues a clip after its end or leads into its start from the clip's own frames, so the motion carries on. The longer clip is a new video beside the source. Slow it down, steady it or track something in it from the same place.
There are more than sixty. Here are eight; the full list is grouped by what they save you from.






















TinyPine ships a fixed set of models and every part of the app is written for how each of them behaves. There are no versions to match, no nodes to wire and no settings to hunt for.
Assist is built on Qwen3-VL, the model family Krea 2 reads prompts with, and writes in the order Krea 2 follows best: the subject first, then style, light and setting.
Sampler, schedule and shift are written for Krea 2's flow-matching schedule, with their own defaults for Quality and Fast and one click back to them.
Train on Krea 2 Raw, the undistilled base, and the LoRA is applied on Turbo, the model you render with. You pick the checkpoint from its sample.
Extend gives LTX-2.5 the clip's own frames as the start of the next part, so the motion carries on. The longer clip is a new video beside the source.
The app keeps each song's codes, so the model can write more of a song later. Artist names in a style are replaced by their genres.
Speech is made one sentence at a time, so one line can be redone. Whisper, with a Swedish model, turns what is said into subtitles.
One click selects an object in Design and in masks; on video it follows the object frame by frame. Cut out lifts a subject with soft edges.
Generate a design, an element or a background on the canvas, or split a flat graphic into separate layers you can edit.
Each model installs from a signed catalog in the precision that suits your card, hash-checked as it downloads, the moment a feature needs it.
One download for Windows. The Grove installs each model when you first use it, sized to your GPU.
Prompts, pictures, video, sound, LoRAs and your library never leave your computer. The app sends a version check, a notice check and an anonymous usage ping you can switch off. Everything it sends
The Grove catalog is signed with Ed25519 and checked before a model installs. Files are hash-checked as they download.
Prompts for sexual content involving minors, or intimate images of real people without consent, are refused. Adult art is not. The policy
Hover a picture to see what it was made from. The whole gallery has every one.
























TinyPine is in active development, so expect frequent updates and a few rough edges. The app tells you when a new version is out. It is free, and version 1 stays free.