ComputeMint
Sign in

ComputeMint · Bengaluru

From files
to something useful.

Bring your Python project for GPU execution. Or turn recordings into text and subtitles.

Explore the services See an output example

Checking the RTX 5090…

Example jobIllustrative example

Program log train.py

$ python train.py --config lora.yaml
Loading the base model in 4-bit (bitsandbytes)
Adding LoRA adapters (PEFT)
Training epoch 1 of 3
Saved checkpoint-500/
Training epoch 2 of 3
Training epoch 3 of 3
Saving the final adapter
Done.

Your results Download everything (.zip)

  • adapter_model.safetensorsadapter
  • adapter_config.jsonconfig
  • checkpoint-500/checkpoint
  • train_log.txtlog

An illustration of a job's log and output files. It is not a real run, a measured result or a benchmark.

The service collection

Useful work.
A considered collection.

Choose by what you want to make.
Read the limits before submitting.

01 / TRAINEarly access

Train and fine-tune.
Your own models.

Run your training project or fine-tune a model with your own code and data.

Bring
Your PyTorch project, data and any pretrained weights (jobs have no internet)
Take away
Checkpoints, adapters, logs and anything your script saves

PyTorch, Transformers, PEFT, bitsandbytes and Diffusers preinstalled. We switch on accounts by hand while we’re small.

See what’s installed
02 / RUNEarly access

Run your models.
Collect the results.

Run batch inference or evaluation with your script, inputs and model weights, then download what your program produces.

Bring
Your script, inputs and the model weights
Take away
Generated outputs, scores and logs

Batch jobs, not a live chat: your script runs, saves results and ends.

How a job runs

Also available: lecture and interview transcription with subtitles

UPSCALING / NEXTPlanned · not available yet

Sharper images
and frames.

Upscale images and video frames with an established open model. Next on the list.

IMAGE / PLANNEDPlanned · not available yet

From a prompt
to an image.

Create or edit images from a prompt or an uploaded image. Planned, not available yet.

3D / PLANNEDPlanned · not available yet

From an image
to a 3D model.

Turn a photo or render into a mesh you can edit or 3D-print.

NOTES / PLANNEDPlanned · not available yet

From your notes
to a narrated video.

Turn study notes into an explanatory video with narration. Planned, not available yet.

VIDEO / PLANNEDPlanned · not available yet

Short clips
from an image.

Short video clips from a prompt or an image, with labelled outputs. This is the last service in the planned build order.

The workflow

Prepare. Review. Run. Collect.

Keep your own copies.
No promised turnaround.

  1. 01 / Choose

    Start with your outcome

    Read the service guide and check your files are supported.

  2. 02 / Prepare

    Make a draft

    Set options and upload inputs. A draft does not spend credits.

  3. 03 / Review

    Check before submitting

    Review the credit charge, the limits and whether the GPU is free.

  4. 04 / Collect

    Inspect your results

    When a job produces results, preview supported files and download them before expiry.

A small start. Room to grow.

More possibilities.
The same care.

Upscaling, image generation and editing, image → 3D, notes → narrated video and video generation are planned, in that order. They are not available yet, and no launch dates are set. Blender rendering and independent providers remain ideas. Exploratory only, not available.

Explore the future direction

Before you start

Clear expectations.
Better decisions.

Which services exist in the preview?

Training and running your own Python/PyTorch projects (script access is enabled per account; ask us) and audio transcription. Jobs have no internet, so bring your data and any pretrained weights in your upload. Upscaling, image generation and editing, image → 3D, notes → narrated video and video generation are planned but not available yet (no dates). Blender rendering and independent providers remain ideas. Exploratory only, not available.

Can I buy a service now?

Real-money payment readiness is not established by this design. Check the app's payment-mode notice and contact us before relying on the local preview for paid work.

Will my job finish at a certain time?

No completion time or successful result is guaranteed. Inputs, configured limits, queues and local-machine availability affect the outcome.

Are the animations real processing activity?

No. Motion is decorative, and the output example is illustrative. Neither is live job status or measured performance.

Have a recording or a project in mind? Tell us what you need before relying on the preview for important work.

Talk to us