Datapoint research grants
$2,000,000 in human preference data for researchers.
If you are working on image, audio, or video models and your research needs real human judgments, we will fund the data. Grants are paid in Datapoint credits and applications are a single email.
At a glance
- Total pool
- $2,000,000
- Typical award
- $5K – $50K
- Decision
- Within 24 hours
- Deadline
- None, rolling review
- Paid as
- Datapoint credits
- Modalities
- Image · Audio · Video
Preference data where generative models are hardest to evaluate.
Image
- Side-by-side preferences between text-to-image models
- Prompt faithfulness and aesthetic quality ratings
- A/B tests on edits, styles, and safety filters
Audio
- Naturalness ratings for TTS and voice cloning
- Preference pairs for music and sound generation
- Speaker similarity and pronunciation judgments
Video
- Side-by-sides between text-to-video generations
- Temporal consistency and artifact ratings
- Edit quality and instruction-following comparisons
Built for research, not production pipelines.
If your work is headed for a paper, a benchmark, or an open dataset, you are in the right place. There is no deadline: applications are reviewed on a rolling basis until the pool is allocated.
A strong fit
- PhD students and postdocs
- Faculty and academic labs
- Independent researchers with prior work to point to
- Nonprofit research organizations
Not a fit
- Data collection for a commercial production model
- Projects with no human preference component
- Reselling the data as a commercial product
Building a commercial product? Datapoint still works for you, just not on a grant. Reach out through the main site instead.
From email to live data in days, not semesters.
Step 1: Email your application
One email in the format below. No portal, no accounts, no letters of support.
Step 2: Rolling review
We read every application and reply within 24 hours.
Step 3: Credits land in your account
Accepted grants are funded as Datapoint credits on an account we set up for you.
Step 4: Run your studies
Launch preference tasks from the dashboard, API, or MCP and watch responses stream in.
One email. Eight lines.
Send the format below to akshat@trydatapoint.com. Plain text is perfect. Attachments and links to prior work help but are not required.
Email template
To: akshat@trydatapoint.com
Subject: Data Grant Application: {your name}, {affiliation}
Name:
Affiliation and role:
Modality (image / audio / video):
Project summary (3-5 sentences):
Preference data needed (task type and scale):
Requested grant size (USD in credits):
Timeline:
Links (paper, lab page, GitHub, Scholar):The button opens your email app with the template prefilled.
Common questions
from researchers
Real human preference judgments on the Datapoint platform: pairwise comparisons, rankings, ratings, and free-text feedback across image, audio, and video. Credits work everywhere: dashboard, API, and MCP.
No. Applications are reviewed on a rolling basis until the $2M pool is fully allocated, and you will hear back within 24 hours of applying.
No. The data you collect is yours, and so is the paper. We ask that you acknowledge Datapoint in publications that use granted data, and we love it when grantees release datasets openly.
It depends on task complexity and annotator targeting, but typical grants fund tens of thousands to hundreds of thousands of judgments. Include the scale you need in your application and we will size the grant with you.
Email us anyway. The grant program is aimed at academic and independent research, but if your work is headed for publication we will figure out the right arrangement.
One active grant per project. Once your study wraps, you are welcome to reapply for follow-up work, and prior grantees who shipped results get priority.
Your models need human judgment. We are funding it.
Eight lines in an email is all it takes, and you will hear back within 24 hours.